E(eit) = 0 for all i, t (69a)

 E.. t = s and all i, j
 E(eit e) = { 'J (69b)
 i s 0 t f s and all i, j


 The stochastic specifications given in equations (69a) and (69b)

serve to characterize the price equations as a system of seemingly un-

related regression equations. Zellner (1962) has shown that the best

estimator for this type of equation system in terms of relative effi-

ciency is a two-stage Aitken's estimator. This estimator was utilized

in estimating the price equation parameters. Before proceeding with a

presentation of this estimator, it is convenient to place the price

equations in matrix form. Let P be an NT x 1 vector of prices, Zi be

T x ki matrix of exogenous variables corresponding to the independent

variables given in equations (67) and (68) for the ith region and e be

an NT x 1 vector of disturbances. The price equations in matrix form

are then given by

 P = Zy + e (70)

where Z is a NT x (zKi) block diagonal matrix with Zi i = 1, ..., N

constituting the diagonal blocks. Further, E(e) = 0 and E(e e') = Qo IT

where IT is a T x T identity matrix and Q has the form


 a11 021 ." alN

 n = 21 22 ". 02N (71)


 N1 aN2 aNN


The Aitken's estimator for y is then given by