FISHERY BULLETIN: VOL. 78, NO. 1 



is adaptive, as the prior distribution is updated 

 each period. Moreover, it is not the same as fixing 

 9 at its estimated value, and using a fixed value of 

 6 in step 3. The difference can be seen in the 

 integral in Equation (4.1). The reason for consider- 

 ing Bayes equivalent policies is that van Hee 

 (1977a, theorem 3.1) proved that for the models 

 under discussion, when the objective is given by 

 Equation (1.2a) or (1.2b), then the Bayes equiva- 

 lent policy is optimal for the full Bayes model. For 

 example, in Walters and Hilborn (1976), the 



parameter 6 is a scalar, i.e., flg ^^ our notation. 

 Their problem, for which an optimal policy was not 

 found, can be solved by following a policy outlined 

 in the five steps above. 



Many models will not have the necessary struc- 

 ture for a Bayes equivalent policy to be optimal for 

 the full Bayes model, and unlike salmon manage- 

 ment, estimates of the population size may not be 

 available every year. A legitimate question is: 

 suppose the present best estimate of B were to be 

 used from hereafter. What would be the loss in 



