to take throiigh the segment for listing 

 purposes. These lists were then returned 

 to the Philadelphia office of the A. J. 

 Wood Research Corporation where they were 

 checked. Next, the dwelling \inits on each 

 list which were associated with the se- 

 lected sampling units were marked for inter- 

 viewing. For example, if a given area seg- 

 ment was assigned three sampling xinlts and 

 the random selection had designated the 

 second sampling unit, the list was first 

 divided into three equal parts and then the 

 dwelling units listed in the second of the 

 three parts were marked for interview. 

 The few sampling units in each urbanized 

 area which contained more than 12 house- 

 holds selected for interview were sub- 

 sampled. The lists were then returned to 

 the interviewers for interviewing. 



The interviewers were instructed to 

 interview the sample (marked) households 

 on the list and any other household (not 

 shown on the list) found between a sample 

 household and the next one listed. TbMS 

 households irtiich might have been emitted 

 in the pre-listing were included; and 

 changes occvirrlng after the pre-listing 

 were accounted for. Interviews in the 

 sample households were conducted with the 

 person mainly responsible for planning the 

 meals. Where the person designated for 

 interview was not at heme on the first 

 call, succeeding calls up to a total of 

 three were made on different days or 

 evenings (in seme instances more than 

 three calls were made ) . 



Tbe sample design for the Orangeburg 

 County, South Carolina, sample was similar 

 in many respects. After preparing an 

 ordered list of the enumeration districts 

 faLLling in the rural portion of the county, 

 sampling units were assigned to the enu- 

 meration districts according to the number 

 of dwelling units occupied in 1950 by non- 

 white household* contained in each. These 

 sampling units were then grouped into geo- 

 graphic strata, 21 in total, with each 

 stratum contsuining 55 sampling units. Two 

 sampling units were selected at lemdom 

 without replacement from each stratum, 

 yielding a total of 42 sampling units for 

 the sample. Next, maps of each of the area 

 segments containing a selected sampling 

 unit were prepared and the interviewers 

 listed all dwelling xuiits falling within 

 the area segment, classifying these 



3l^ 



dwelling units according to whether they 

 were occupied by white households or non- 

 white households or were vacant. The 

 location of each dwelling unit was marked 

 on the segment map and numbered; this same 

 number was used on the listing sheet. 



Field Work 



Training sessions with the super- 

 visors and interviewers were conducted in 

 each of the survey areas by members of the 

 Philadelphia office staff of the A. J. 

 Wood Research Corporation. Initial field 

 work was checked for quality and under- 

 stemdlng of the instructions. 



In addition to the check of the 

 initial interviews, the area supervisors 

 were required to conduct a preliminary 

 edit of all work turned in and to check 

 10 percent of each interviewer's work by 

 telephone. A further verification check 

 on the field staff was carried out by the 

 home office by means of a check card mail- 

 ing to 33 percent of the respondents in 

 each city. 



A total of 2,385 households were 

 designated for interview in this survey; 

 706 in Birmingham, 7^*3 in Boston, 716 in 

 Detroit and 220 in Oreuigeburg. Interviews 

 were completed in 1,9*^7 of tiie sample 

 households. The reasons for the non-inter- 

 views are tabulated in Appendix Table 2. 



Data Processing Procedures 



All questionnaires were edited upon 

 receipt in the Philadelphia office and 

 those which were incomplete or contained 

 questionable responses were returned to 

 the field supervisors for re-interview. 

 The coding department then prepared tabu- 

 lations of the open-end questions from a 

 sample of the completed Interviews from 

 each survey area. Codes for these ques- 

 tions were established and coding instruc- 

 tions prepared and reproduced. 



The questionnaire and coding proce- 

 dures were explained and reviewed with 

 the coders. Tbe open-end questions were 

 reserved for coding by the most experi- 

 enced coders only. Answers to open-end 

 questions ^Ich were not readily 



