X- E5.3.3 Multicollinearity
If there is a linear combination of the independent variables which produces a column of zeros – i.e., at least one column of X can be represented as a linear combination of other columns of X then the least squares regression coefficient cannot be computed by inverting the moment matrix. In this case, the minimizer of the sum of squared residuals is not unique. As a general rule, LIMDEP does not proceed any further if it detects that your data are collinear. (Some other programs will successively drop variables from the equation until a noncollinear set remains, as if to report that your desired model was inestimable , so the program found some other model that was. While users differ in their preference for this kind of program driven specification, LIMDEP adheres to a strict rule of always waiting for the user to specify the model to be estimated.)