p.866

Single Equation Vs. Equation System Methods of Estimation in Econometrics

L. R. Klein

Abstract

The choice between single-equation and equation-systems methods of estimation is often based on improper criteria and unjustified claims. On the one hand, advocates of the equation-systems approach probably overstated their case in the early years of enthusiastic development. While the gains to be realized from the use of more powerful statistical methods are quite modest, they are, nonetheless, real and should not be neglected. On the other hand, coefficient-by-coefficient comparisons of structural estimates by the two approaches are not suitable in many situations. There is need for comparisons of summary statistics of a whole system. In particular, reduced form coefficients may show large differences even though component structural coefficients appear to be close. In addition, the efficiency properties of single equation least squares estimates do not hold under transformation from structural to reduced form coefficients. These points are clearly revealed in a number of recent Monte Carlo studies.