Seemingly Uncorrelated Regressions

This chapter considers the estimation problem of the complete system. It also considers various estimation techniques for the parameters in Equation. There are two ways to estimate System: Estimating each equation by taking into account the identifiability restrictions on only that equation; estimating all equations together by taking into account all the identifiability restrictions in the system. In the former case the estimators are called limited information (LI) estimators whereas in the latter case they are called full information (Fl) estimators. The chapter outlines the procedure adopted by Nagar for deriving the bias and moment matrix of the general k-class estimator and state the results obtained. It shows s necessary and sufficient condition for the identifiability of ₁ is the same as in the classical case. However, in some special cases like recursive systems or systems with no feedback among endogenous variables it was noted that the problem of identification reduces to that of classical cases.