to the equality of the reduced form (and structural) parameters across

states. It should be noted, however, that the mean of each region's

stochastic process, Ai, is unconstrained across equations. The specifi-

cation and intrepretation of the Ai parameters, in particular, and the

stochastic processes characterizing the regression disturbance, in

general, provide a convenient introduction to the discussion concerning

the choice of the appropriate estimator.

 The geographic location of the primary reef fish stocks are

depicted in Figure 10. There is some indication, although there is not

an overwhelming amount of evidence, that the reef fish stocks do not

exhibit a great deal of migratory behavior (GMFMC, 1979). This fact,

taken in conjunction with the large geographic dispersion of fishing

grounds, suggests that the GMRFF is composed of several biologically

independent stocks of reef fish. Additional information on the general

fishing locations of vessels originating from various states' ports

(GMFMC, 1979), indicates that vessels originating from different states

fish on common grounds. This information suggests each state catch

function should have a stochastic process dominated by the stock most

frequently fished, and that these processes should be contemporaneously

correlated due to the intermixing of vessels from different states.

Thus, the overall structure of the system of catch equations is char-

acterized by a system of seemingly unrelated regression equations (SUR)

with cross equation parameter restrictions and autoregressive

disturbances.

 The information above also serves to give an interesting interpre-

 tation to the Ai parameters. These parameters serve to determine the

 "location" of the catch equations for each state in input-output space.