1. Grouping the eligible population 

 into snail clusters or sampling 

 units conprising an efficient 

 interviewer daily work load. 



2. Gro'ipinfT the sampling units 

 into city and suburban zones, 

 in each of the three urbanized 

 areas surveyed, to provide 

 approximately proportionate 

 representation. 



3 . ji f urtlier grouping of the 

 sampling units Kithin each 

 zone into geographic or area 

 strata, xirith an equal number 

 of sar.pling units in each 

 stratum, to ensure adequate 

 distribution of the saFiple to 

 all segments of the population 

 of interest. 



h. Using equal probabilities for 

 the selecting of sampling units 

 within strata and thereby con- 

 siderably simplifying the formu- 

 las necessary for valid compu- 

 tation of the estimates and of 

 their standard errors. 



A strict probability sample implies 

 the application of completely objective 

 methods for the selection of respondents. 

 In the absence of a list of households or 

 persons eligible for interview, the re- 

 ouired objectivity is met through the use 

 of area probability sampling technioucs. 

 To be satisfied with simple area sa;"plinf^ 

 techniques is not enough, however. Inge- 

 nuity in the use of available resources 

 and facilities can considerably increase 

 the efficiency of one area probability 

 sample over another. 



United States P-^isus Population end 

 Housing data, both published and unpub- 

 lished, are our major resource in the 

 design of efficient probability sam.ples. 

 Unpublished data for small areas, such as 

 enurr.erat ion districts used in collecting 

 census data, may be purchased on special 

 order from the Bureau of the Census, In 

 open countrjr areas maps indicating the 

 location of dwelling units are available 

 from State Highway Commissions, This 

 supplementary- information may be used for 

 a variety of purposes in the design of a 

 sample including stratification, assigning 

 selection probabilities, or for the con- 



struction of approximately equal-sized 

 sampling units. The san.ple design out- 

 lined below makes use of 1950 census data 

 to establish the area strata and for the 

 assignment of the sampling units within 

 these strata. Although these data were not 

 used for the direct assignm.ent of selection 

 probabilities , the sajnpling plan adopted 

 is such that the chance for any segment of 

 the areas surveyed, to be represented in the 

 sample, was approximately proportionate to 

 the number of occupied dwelling units 

 contained within the segment whether it was 

 an enumeration district, census tract, 

 township, urban place, city block, or 

 portion of an enujrierat ion district, etc. 



/I sample representative of all house- 

 holds in the urbanized areas of Birmingham, 

 /ilabama; Boston, Massachusetts; Detroit, 

 Michigan was selected for this study. In 

 addition, a representative sample of all 

 non-white households located in the rural 

 portion of Orangeburg County, South 

 Carolina was chosen. Bureau of the Census 

 definitions of households, di-;elling units, 

 urbanized area, rural territory, etc. v/ere 

 employed. The sample designs for the 

 three urbanized areas will be described 

 first. These designs were stratified one- 

 stage sample designs, constructed in 

 accordance with the principles outlined 

 above. Careful control in all steps of 

 the sample selection made it possible to 

 know exactly the chance every household 

 cluster or sampling unit had of falling 

 into the sample. 



The first step in the sairiple design 

 consisted of listing and ordering geo- 

 graphically the census tracts in the 

 central city portions of each of the three 

 urbanized areas. In Detroit, those census 

 tracts with 10 percent or more of the 

 cfcjelling units occupied by non-white house- 

 holds in 1950 were listed and ordered 

 separately. Similarly, ordered lists of the 

 1950 Census Jliiumerat ion Districts were 

 prepared for those portions of the three 

 urbanized areas which fall outside the 

 central cities. Geographic or area strata 

 were then constructed within the central 

 city zones and the suburban zones for each 

 of the urbanized areas using the ordered 

 lists and 1950 census data an the number of 

 occupied dt/elling units or households found 

 in each census tract, block or enumeration 

 district. These strata, seventy in number 

 for each urbanized area, vjere constructed 



Uh 



