MENDELSSOHN. USING MARKOV DECISION MODELS 



creases the percent of time there are low catches, 

 and only reduces the mean per period harvest by 

 33,800 fish. In order to achieve a smoother catch, 

 "potlatch" harvests from time to time have been 

 sacrificed. 



When looked at closely, these policies are actu- 

 ally very intuitive and represent an interesting 

 variant of a base stock policy. These policies re- 

 place a single base stock size by a dual base stock 

 size policy. The first base stock size is lower than 

 the original one, while the second base stock size is 



greater than or equal to the original base stock 

 size. This means that there are fewer states where 

 there is no harvesting, but also lowers the likeli- 

 hood of the really big harvests. The mean per period 

 harvest tends to be very sensitive to these big 

 harvests, while the median is not, particularly 

 since the very large harvests are not too frequent. 

 It is curious that the population dynamics are so 

 sensitive to such fine tuning, for the difference 

 between policy 1 and policy 3, say, is quite mar- 

 ginal. It would be an interesting area of future 



.56 112 1.68 224 2.80 336 3.92 4.48 5.04 5.60 6.16 6.72 .56 1.12 1.68 2.24 2.80 3.36 3.92 4.48 504 5.60 6.16 6.72 



X X 



Figure 2.— Continued. 



43 



