mechanistic explanation as to why these variables 

 should affect the catch and effort, they are not 

 being used at this time in the forecasts. (How- 

 ever, the ability to include random environmental 

 factors into the forecasting model is an advantage 

 when using stochastic models as compared with 

 the normal deterministic production models.) Dis- 

 aggregating by size class might also improve the 

 forecasts. Prior to 1973, the catch of the large 

 skipjack tuna and the total catch were highly 

 correlated. Since 1973, this has not been true and 

 there has been a definite change in the size 

 composition of the catch. A disaggregated inter- 

 vention model may be able to explain this change. 



SUMMARY 



Box-Jenkins models have been proposed as 

 an alternate model for forecasting fishery data. 

 ARIMA models provide maximum likelihood esti- 

 mators that are not biased when the data is 

 seasonal and autocorrelated, and when a variable 

 is lagged on itself. Techniques are explored which 

 allow the model to be constructed from the data 

 up, rather than from theoretical models that may 

 not be supported by the data. The procedure is 

 illustrated on skipjack tuna catches in Hawaii, 

 which traditionally has been considered too 

 variable to forecast on a monthly basis in a 

 reasonable manner. 



ACKNOWLEDGMENTS 



I am indebted to Lisa Katekaru for her invalu- 

 able assistance in performing most of the pro- 

 gramming for this paper. Robin Allen and two 

 anonymous referees provided several comments 

 that have improved the presentation and discus- 

 sion of this paper. 



FISHERY BULLETIN: VOL 78, NO. 4 



LITERATURE CITED 



ABRAHAM, B., AND G. E. P. BOX. 



1978. Deterministic and forecast-adaptive time- 

 dependent models. Appl. Stat. 27:120-130. 

 ANDER.SON.O. D. 



1975. Time series analysis and forecasting: The Box- 

 Jenkins approach. Butterworths, Lond., 182 p. 



ATCHLEY, W. R., C. T GASKINS, AND D. ANDERSON. 



1976. Statistical properties of ratios: I. Empirical re- 

 sults. Syst. Zool. 25:137-148. 



Box, G. E. P., AND G. M. JENKINS. 



1976. Time series analysis: forecasting and control. Rev. 

 ed. Holden-Day, San Franc, 575 p. 



CHAYES, F. 



1949. On ratio correlation in petrography. J. Geol. 57: 

 239-254. 

 EBERHARDT, L. L. 



1970. Correlation, regression, and density dependence. 

 Ecology 51:306-310. 

 Fox, WW, JR. 



1970. An exponential surplus-yield model for optimizing 

 exploited fish populations. Trans. Am. Fish. Soc. 99: 

 80-88. 



1971. Random variability and parameter estimation for 

 the generalized production model. Fish. Bull., U.S. 69: 

 569-580. 



1975. Fitting the generalized stock production model by 

 least-squares and equilibrium approximation. Fish. 

 Bull., U.S. 73:23-37. 



Granger, C. W J., and P Newbold. 



1977. Forecasting economic time series. Acad. Press, 

 N.Y.,333p. 



Johnston, J. 



1972. Econometric methods. 2d ed. McGraw-Hill, N.Y., 

 437 p. 



Newbold, R, and n. Davies. 



1978. Error mis-specification and spurious regressions. 

 Int. Econ. Rev 19:513-519. 



Parrish, r. h.. and a. D. MacCall. 



1978. Climatic variation and exploitation in the Pacific 

 mackerel fishery. Calif. Dep. Fish Game, Fish Bull. 167, 

 110 p. 



PELLA, J. J., AND R K. TOMLINSON. 



1969. A generalized stock production model. Inter-Am. 

 Trop. lYina Comm., Bull. 13:419-496. 



896 



