of instances this information was refused. 

 In other instances it was found that the 

 records of the food facilities were such 

 that no accurate data could be obtained. 

 as a result of these two situations, form 

 "R" schedules were completed with 352 

 feeding facilities, in 350 plants. In two 

 plants, two separate feeding facilities 

 were audited. The 350 plants surveyed * 

 represented a total of 361 establishments . 



Naturally, not all of the food sched- 

 ules covered precisely the same period in 

 time. For the typical or "median" facility, 

 the initial inventory was taken January 10, 

 1956 and the closing inventory February 8, 

 1956. All food purchases of this period, 

 approximately four calendar weeks, were 

 recorded and food consumption calculated. 



The average period covered for all 

 facilities surveyed was 28.7 days. (These 

 are calendar days and the number of working 

 days covered is affected not only by the 

 number of week-ends but also by the extent 

 to which the plant may operate on a 5-1/2 

 or 6 day week.) A distribution of the num- 

 ber of calendar days covered in these 

 schedules is as follows: 



Less than 27 days .... 6.6$ 



27 days 32.5 



28 days 18.5 



29 days 9.8 



30 days 15. h 



31 days 7.7 



32 or 33 days 6.5 



More than 33 days .... 3.0 



'//eight ing System 



The companies with feeding facilities 

 with which "R" questionnaires were com- 

 pleted were drawn from those companies 

 which had been identified during telephone 

 screening operations as having feeding 

 facilities. 



Also as mentioned earlier, the basic 

 sample itself was disproportionate both 

 by size and by region, and a weighting 

 system was found necessary to restore to 

 this sample the proportionality which 

 existed in the universe. That is, the "T" 

 interviews were stratified by size and 

 region, and such weights assigned to each 

 region-employee -size cell as would effect 

 this restoration. This same need for 

 weights extended to interviews made with 



sub-samples. Weights were also required 

 to take into account the effect of re- 

 fusals to cooperate during the survey. 

 The 391 "A" plants, the 378 "B" plants, 

 and the 350 "R" plants must all necessar- 

 ily be considered samples drawn from the 

 same universe (although biases are intro- 

 duced by the failures to cooperate) and 

 suitable weightings are needed in order 

 to make valid comparisons of materials 

 obtained in one questionnaire with 

 material obtained during another. 



The universe, as will be remembered, 

 was originally defined in terms of "es- 

 tablishments" and in terms of the size of 

 these establishments as this was reported 

 to be BOAS I in 1953. Since the sample 

 was selected on this basis weights must 

 also be calculated and applied on the 

 same basis. However, it should be spec- 

 fically noted that while weights are 

 calculated on the basis of 1953 size of 

 establishment, tabulations have been 

 presented in terms of 1956 size of plant. 



Generally speaking this procedure was 

 followed: The universe with which the 

 sample is associated was derived either 

 from BOASI statistics, or, in the case of 

 the universes of "plants with feeding 

 facilities" calculated on the basis of 

 telephone interviews. Then, a proportional 

 sample was devised on a twelve-cell "geo- 

 graphic area by 1953 plant size" basis, so 

 that the number of interviews in this pro- 

 portional sample in each cell was a con- 

 stant fraction of the universe of the es- 

 tablishments in the cell. The "propor- 

 tionate sample" in each cell was then 

 divided by the actual number of establish- 

 ments interviewed in that cell. The re- 

 sult, extended two decimal places was the 

 weight assigned to all establishments in 

 the cell. 



Since the sample plants were drawn 

 from the same universe and in the same 

 manner as the sample of "establishments", 

 the same weights were applied to both. 



Basis For Universe Projections 



Data from the sample on food consump- 

 tion in in-plant food services have been 

 projected to indicate the magnitude of the 

 total market for food in such facilities. 



Naturally, these projections have 



30 



