The field work of interviewing the ovmers of the New England 

 vessels in the second-stage sample was conducted in Kay of 1956, earlier 

 than in the Gulf Area and California, Field work supervisors were trained 

 by interviewing vessel owners in Boston. Subsequently each field work 

 supervisor was assigned to a Ke;-j England port to obtain and train inter- 

 viewers, supervise the interviewing, conduct local research, and submit a 

 report of his findings. The interviev;ing of vessel owners in the Gulf Area 

 and in the California Area began early in Jime of 1956 \mder the direction 

 of the same supervisors who carried out the field work in New England. Late 

 in June the study of insurance files of the New England vessels in the third- 

 stage sample was begun. Field work supei'visors were trained with the insurance 

 records available at Boston insiirers and insurance agents. VJith the aid of 

 locally recrxiited assistants, these field work supei'visors then continued to 

 collect information on insurance experience from no less than 50 insurance 

 firms and agents located in Gloucester and Nev; Bedford, Massachusetts; New 

 York, New York; Philadelphia, Pennsj-lvaniaj New Orleans, Louisiana; Houston, 

 Texas; Los Angeles, San Diego and San Francisco, California, 



Early field work in the New England Area offered a number of 

 advantages. Reliable, competent, and well trained personnel were given the 

 responsibility of supervising field work in the other areas, Interviev; 

 schedules and instructions were tested. Finally, cost estimates for the whole 

 study were made in the light of difficulties enco\mtered in New England. 



U. Sample validation . 



A number of correlations and chi-square analyses were made to test 

 the following: (a) whether fir^t-stage sampling has affected significantly 

 the representativeness of the second-stage samples; (b) the assumption that 

 the stratifying variables are correlated with the subject under study — insurance 

 experience; (c) the representativeness of the second-stage samples; and (d) 

 the representativeness of the third-stage samples, 



(a) First-stage samples. The exclusion of a number of home 

 ports in New England and the Gulf Area through first-stage 

 sampling does not seem to have affected significantly the 

 representativeness of the second-stage samples. First, the 

 findings indicate that during 19S0-Sh the vessels in the second- 

 stage san^iles had used for fish landings a large number of ports 

 in each area (26 in New England, 39 in Gulf, lIi in California), 

 including the home ports which vjere eliminated through first- 

 stage sampling. Second, stratification of the second-stage 

 samples was based on the stratifying characteristics of the initial 

 universe, not of the first-stage sample. Finally, chi-square analysis 

 of the second-stage samples leads to the sarne conclusion, (See 

 table B-8 of this appendix,) 



(b) Correlation of stratifying variables . The assumption that 

 the stratifying variables are correlated with insurance experience 

 is substantiated by the findings. First, the average insurance 

 rate for active vessels is lovjer than the rate for lost vessels by 



321 



