A recursive two-step least squares method

A two-step method of estimating the parameters of a regression equation on a finite sample is discussed. Conditions are obtained under which the simpler-to-compute two-step estimates feature a smaller r.m.s. error that the errors of the method of least squares. These conditions are shown to represent situations where identification's difficult. Important statistical characteristics of two-step estimates are obtained. The results of computer testing of the two-step method are reported.