HUPPERT: ANALYSIS OF UNITED STATES DEMAND FOR FISH MEAL 



coefficients satisfy prior expectations. Because it 

 yields a significantly higher r^, and because the 

 test for serial correlation suggested by Griliches 

 (1967) lends it support, I tend to favor the distrib- 

 uted lag model. But the evidence is not really 

 conclusive. For one thing, the "Griliches test" 

 looks only for first-order serial correlation, and it 

 will probably fail to give correct guidance when 

 more complex residual generating processes are 

 present. Another difficulty is the lower precision 

 of the regression coefficients in the distributed 

 lag model. The importance of this depends upon 

 how the demand function is to be used. In 

 fisheries management applications the most im- 

 portant use of the demand model will be for pre- 

 dicting price effects resulting from changes in 

 annual production. 



To compare the two demand models, the equa- 

 tions are transformed to give quantity demanded 

 in natural units (tons offish meal proteins) and 

 the 1976 values of independent variables other 

 than fish meal price are inserted. The resulting 

 relationships between price and quantity are 



Table 7. — Demand predictions (q) and price elasticities {E) for 

 static demand (X = -0.55) and partial adjustment (\ = -0.3) 

 models. 



for the dynamic demand model. Quantities pre- 

 dicted by Equations (15) and (16) and price elas- 

 ticities of demand for a range of prices are listed 

 in Table 7. From the Table and Figure 1 it is clear 

 that the two demand models are grossly simi- 

 lar. At low supply levels (less than about 250 t), 

 however, the predicted price responses are 

 greatly different, as are the quantities demanded 

 when prices are low (<$4 per unit protein). Thus, 

 any conclusions reached on the basis of this de- 

 mand analysis will be sensitive to the specification 

 of the demand function. 



Qt = 



-0.00389 + 0.04916 



fp —0.55 



-0.55 



—0.55 



(16) 



for the static demand model, and 



Qt 



_, 1 



—0.3 



-0.03486 + 0.18001 



-0.3 



-0.3 



(17) 



Figure l. — Fish meal demand curves 

 based upon the maximum likelihood es- 

 timates of the static demand model (X = 

 -0.55) and the partial adjustment 

 model (X = -0.3). 



100 



200 300 400 500 600 700 800 



FISH MEAL PROTEIN (1,000 metric tons) 



900 



1000 



275 



