Appendix 



SAMPLING AND SURVEY METHODOLOGY 



The Universe From Which The Sample Was 

 Selected 



In general, the universe from which 

 the sample was selected (for the entire 

 series of publications) consisted of 

 28,Lb6 manufacturing establishments which 

 had 100 or more employees during the first 

 quarter, 1953, according to the records of 

 the Bureau of Old Age and Survivors 

 Insurance (BOjiSI). Establishments which 

 had fewer than 100 employees in 1950 or 

 which were not in existence as of that 

 date, and all non-manufacturing establish- 

 ments, had a zero probability of inclusion 

 in the sample, and are therefore not repre- 

 sented in the survey findings. 



For the purposes of the "commodity 

 reports", the survey was restricted to 

 larger establishments, i. e., establish- 

 ments with 2^0 or more employees in first 

 quarter, 1953. Although basically the 

 same procedure was used for sampling all 

 establishments, certain refinements were 

 introduced to improve the efficiency of 

 the sample of establishments with 250 or 

 more employees. 



Design of the "Large Plant" Sample 



The sampling method used for the 

 "large plant" sample may be described as a 

 self -weighting cluster sample, with clus- 

 ters chosen for inclusion in the sample 

 with probability proportional to size. 



Prior to the selection of survey 

 respondents, primary sampling units (psu's) 

 were formed. Each psu was comprised of a 

 single county, or a group of adjacent 

 counties, with a minimum population of 

 fifteen manufacturing establishments with 

 250 or more employees. Excluded from the 

 universe (and from the sample) were nine- 

 teen "statewide" establishments; those 

 with no fixed place of operation and which 

 could not therefore be associated with a 

 specific psu. 



In total, 267 primary sampling units 

 were formed, with a total population of 

 11,1*01; establishments. The remaining U70 

 establishments were in "unclustered 



counties" and were placed in a separate 

 stratum. (A discrepancy of one establish- 

 ment in the published B0ASI statistics 

 was ignored.) 



The psu's and unclustered counties 

 were stratified by four geographic re- 

 gions. The number of psu's in each re- 

 gion, the population of these psu's, and 

 the number of establishments in "un- 

 clustered counties" is as shown in the 

 following: 



Area 



Number Total 

 of estab- 

 psu's lishments 

 formed in psu's 



Total 

 establish- 

 ments in 

 unclustered 

 counties 



Northeast 75 U,l6l 



North Central 92 1*,067 107 



South 81 2,353 280 



West 19 823 83 



All regions 267 ll,l+0li 



1*70 



The probability of the inclusion of 

 any psu in the sample was proportional to 

 the size of the psu measured in terms of 

 numbers of establishments with 250 or more 

 employees. Separately within each of the 

 four geographic regions, a random start 

 and a sampling interval were designated 

 and a total of 50 primary sampling units 

 were selected. In addition, counties were 

 selected at random from the stratum of 

 unclustered counties. 



Several psu's, because their size 

 exceeded the sampling interval, were 

 selected more than once; this was true of 

 the five counties in New York City (which 

 are treated as a single unit by the B0ASI), 

 the psu made up of Cook and DuPage 

 counties in Illinois, and the psu created 

 from Los Angeles, Ventura, Orange, Kern, 

 and Santa Barbara counties in California. 



The establishments within each psu 

 were divided into three strata on the 

 basis of 1953 employee size: 250 to 1*99 

 employees, 500 to 999 employees, and 

 1,000 or more employees. For each stratum, 

 a random start was selected and an inter- 

 val determined which would provide a 

 self -weighting sample. The actual selec- 

 tion of establishments in sample psu's 



27 



