APPENDIX B 



SAMPLING TECHNIQUES AND PROCEDURE 



1, Initial Plan and Reasons for Their Revision 



The initial plan . Tentative sampling plans in the contract pro- 

 vided for the study of about 1,000 insurance policies and their accident 

 claims for hull insurance and an equal number for protection and indemnity- 

 insurance carried by commercial fishing vessels of ^ net tons or more for 

 each of the tv;o specified years, 19^1 and 195ii. The policies were to be 

 drajm from no less than 10 insurers which have written hull insurance and 

 no less than 5 insurers which have written protection and indemnity insuj^ance. 

 In addition, the contract specified the personal interviewing of about 500 

 owners of commercial fishing vessels of 5 net tons or more for each of the 

 two years already mentioned. The randomly selected samples of policies and 

 owners to be distributed among six major regions of the continental United 

 States, namely New England, Middle Atlantic, Gulf Area (including South 

 Atlantic), California, Pacific Northwest (Washington and Oregon states) and 

 inland v;aters. 



Reasons for revision of the initial plan . Early in the course of 

 our preliminaiy investigation a number of serious obstacles appeared to pre- 

 vent the completion of these contractual specifications to their fullest 

 extent. 



First, it was found that insurers often keep their accouints on 

 commercial fishing vessels mixed with other accounts. Information about the 

 kind of insurance carried by each vessel could not be obtained without a 

 complete preliminary survey of the universe. The cost of obtaining such 

 information about the universe for sampling purposes appeared, according to 

 all estimates, prohibitive. 



Second, officials of insurance companies expressed the firm 

 opinion that no loss experience co\ild be meaningful unless it covered a 

 period of no less than five consecutive years. 



Third, the most obvious interpretation of the contractual speci- 

 fications required three types of samples (hull policies and their accident 

 claims; protection and indemnity policies and their claims, and vessel 

 owners) drawn independently. If sampling had been carried out on this basis, 

 no correlation of data of each sample with data of the two other types of 

 samples woxild have been feasible or easily arranged. 



Fourth, our pi-eliminary inquiry disclosed that the physical 

 characteristics of the commercial fishing vessels, their fishing operations, 

 and other variables important from the viewpoint of insurance experience 

 differed markedly from area to area. 



Fifth, it became evident that insurance experience is likely to 

 be affected sometimes profoundly, by non-qualitative factors. 



317 



