SQUIRES ET AL.: JAPANESE AND U.S. SABLEFISH MARKETS 



Pacific coast ex-vessel prices are also generally 

 lower than the Alaska ex-vessel prices, perhaps 

 reflecting a faster growth rate in market demand 

 for Alaska's fish. An upward trend also exists for 

 all prices. Because the Tokyo prices correspond to 

 eastern dressed weight and correspond to a higher 

 market level, the Tokyo prices can be generally 

 expected to lie above the ex-vessel prices. Inter- 

 estingly, this relationship only begins some time 

 in 1983, perhaps reflecting the increasing impor- 

 tance of the Tokyo wholesale market as a clear- 

 inghouse for the increased imports from the United 

 States and the sharp drop in Japanese harvests that 

 began about this time (so that Tokyo prices now 

 include additional transport, handling, and other 

 market costs). Finally and most importantly, 

 because no simple, direct, one-to-one relationship 

 appears to exist between the Tokyo and U.S. prices, 

 regression analysis can make an essential contribu- 

 tion to understanding the nature of the market 

 integration. 



To apply the Ravallion model, we specified the 

 Tokyo wholesale market as the central market and 

 the Pacific coast and Alaska ex-vessel fixed gear 

 markets as the local markets.^ The unrestricted 

 model given by Equation (1) for six lagged periods 

 (J = 6) was estimated by two-stage least squares. 

 The autocorrelation and partial autocorrelation plots 

 for the residuals were reasonably flat for both local 

 markets, indicating the serial correlation does not 

 present a problem. All statistical tests were F-tests 

 for linear restrictions, which were all evaluated at 

 a 5% level of significance. The results from these 

 F-tests are reported in Table 1. 



The importance of the local market characteristics 

 was first examined. As indicated in Table 1, the null 

 hypothesis that seasonal dummy variables for Alas- 

 ka were not important was not rejected at 5%. Also, 

 the linear trend variable was excluded because it did 

 not contribute to the overall explanatory power of 

 the Alaska model. The 1984 dummy variable was 

 omitted from the Alaska model because 1984 was 



Table 1 .—Hypothesis tests for the integration of Japanese, Pacific 

 coast, and Alaska sablefish markets. Distributions of F-test 

 statistics given in parentheses of form (numerator degrees of free- 

 dom, denominator degrees of freedom). 



^Within-sample, bivariate direct Granger causality tests (see 

 Squires 1986 for a discussion) were first applied to verify if Tokyo 

 is indeed the source of price formulation in the U.S. markets, or 

 whether prices are simultaneously determined between Tokyo and 

 the local markets, or whether price linkages even exist at all. The 

 null hypotheses that prices were first formed in either the Alaska 

 or Pacific coast ex-vessel fixed gear markets were rejected at a 

 5% level of significance in both instances. The null hypothesis that 

 sablefish prices first formed in the Tokyo wholesale market lead 

 the Pacific coast ex-vessel fixed gear sablefish prices was also 

 rejected, thereby suggesting market segmentation. The null 

 hypothesis that Tokyo prices lead Alaska ex-vessel fixed gear prices 

 was not rejected, indicating some form of sablefish market integra- 

 tion with price leadership most likely coming from Tokyo. 



Local market 



Null hypothesis 



Pacific coast 



Alaska 



No local seasonality 

 No local time trend 



and 1984 dummy 

 No local time trend 

 Long-run integration 

 Short-run integration 

 Short-run integration 



(viieak form) 

 Short-run integration 



(weakest form) 

 Market segmentation 



0.458" (4,47) 

 1.429- (2,51) 

 2.870- (1,52) 



0.826- (6,52) 



3.309 (4,17) 



0.034* (1,22) 

 0.023" (1,23) 

 4.230 (12,24) 



4.012 (2,24) 



6.621 (1,24) 



NOTES; Ttie unrestricted model is Equation (1) for J = 6 estimated using 

 two-stage least squares The table gives F-tests of the linear restrictions on 

 the model implied by each null hypothesis Short-run integration tests con- 

 ditional upon maintained hypothesis of long-run integration as given in Equa- 

 tion (9). 



' indicates nonrejection of null hypothesis at 5% level of significance 



the first year that the Alaska's price series was 

 used in the analysis. The seasonal dummy variables 

 did not contribute in a statistically significant 

 way to the unrestricted model for the Pacific coast, 

 so that the quarterly dummy variables were not 

 included in further regressions. The 1984 dummy 

 variable and linear time trend did not contribute 

 to the overall explanatory power of the unrestricted 

 model for the Pacific coast when taken as a group 

 (but not individually). The linear time trend for 

 the Pacific coast was nonetheless retained in the 

 model because of the clear upward trend in real 

 prices exhibited by the data. The final version of 

 the unrestricted model given in Equation (1) does 

 not have any local market characteristics for Alaska, 

 and only includes a linear time trend for the Pacific 

 coast. The regression results for the final versions 

 of the unrestricted model are reported in Table 2. 

 These final versions of the unrestricted model were 

 then used for the hypothesis tests on the form of 

 market integration. 



The next null hypothesis which was tested was 

 that of long-run market integration for Alaska. As 

 indicated in Table 1, it was not rejected at a 5% level 

 of significance. In order to obtain more efficient 

 estimates of the parameters and more powerful 

 statistical tests for the short-run market integration 

 hypothesis test for Alaska, the model was respe- 

 cified with long-run integration imposed as in Equa- 

 tion (9) and all subsequent tests conducted against 

 this restricted form. (The regression results are 

 available from the authors upon request.) All three 



347 



