The codinp instructions were reproduced and training 

 sessions wen. conducted to familiarize coders thoroughly with the 

 questionnaire and with the requirements of coding procedure for 

 the particular job. 



Open-end codes were not finally determined until the 

 field work was completed, ^uestionnsdres were sorted by major 

 geographic areas and systematically sampled within those areas. 

 The numL°r of cases used in the s?mple was 5>00. Separate codes 

 were established for any category which showed up in the sample 

 with a frequency greater than l-gy». This corresponds to reasonable 

 certainty (probability 19 in 20) that any category wid_ch is 3>- or 

 more in the whole population would be separately coded. 



All answers coded into the miscellaneous category were 

 recorded verbatim on a separate card ['or each answer and identified 

 by questionnaire and cluster number, '^his permitted changing the 

 code from a miscellaneous to a specific category if it turned out 

 that this category was much r;ore numerous than the sample indicated. 



The following consistency checks and verification procedure 

 were followed for both straight and open-end codes: 



a. The initial work of individual coders (subsequent to 

 practice sessions) was completely checked by coding supervisors. 

 Any errors found were corrected, necessary changes in the codes or 

 coding instructions were made, and the procedure was completely re- 

 viewed with all the coders. 



b. After coding differences were reduced to an acceptable 

 level, sample verification was instituted. This process is not pri- 

 marily designed to locate errors but to insure that errors and dif- 

 ferences between coders are in control, ^or straight questions a 

 10$ check v/as carried out; for open-end questions the rate was 20^, 



c. Two persons outside the coding section systematically 

 checked the x^ork of the coding verifiers. 



Punched cards were usually prepared while coding was still 

 in progress. A thorough- going series of checks, covering all data 

 on the punched cards, was performed on an electronic statistical 

 machine, following this, a mechanical sequence check was made on all 

 cards to determine that each questionnaire had a complete card file. 



31U 



