Sampling Within Localities . The population in each 

 locality was first divided as between the central city or cities, 

 smaller towns and villages, congested suburban areas and open 

 country areas, and each of these groups was sampled separately. 

 For central city or cities, clusters were defined consisting of 

 combinations of city blocks or in some cases Census Enumeration 

 Districts. One such cluster was selected from each geographic 

 stratum within the city. The smaller cities and towns were first 

 arranged by size and in some instances also by geographic area and 

 one such town or village was selected from each stratum. The 

 selected town or village was further subdivided according to Census 

 Enumeration Districts which were selected as final clusters to be 

 visited by interviewers. For the congested suburban area and open 

 country areas, Census Enumeration Districts were stratified geo- 

 graphically and one district was selected from each stratum with 

 probability proportionate to 1°5>0 population. 



Sampling within Selected Clusters . The clusters consisted 

 of city blocks, or Census Enumeration Districts for which the Census 

 Bureau has provided data on the total number of dwelling units in 

 1950. Host of these clusters range from 1^0 to 2^0 dwelling units. 



Every cluster was prelisted in advance of the final stage 

 of sample selection - the selection of households for interviewing • 

 This means that field workers worked every cluster in advance and 

 made a listing of all dwelling places it contained, showing house- 

 hold addresses and other clear identifications. By applying random 

 numbers to these lists, it was possible to achieve a truly known 

 probability selection of dwellings, and to assign specific addresses 

 for interviewing so as virtually to eliminate chance of interviewers ' 

 choice or judgment influencing the selections. The contractor's 

 New lork City office staff selected the sample households from these 

 clusters and returned the lists to the field workers, showing the 

 specifically designated sample households to be used. 



Although the prelisting was done as an independent operation 

 occupying the full attention of a person specifically trained to do 

 it. the final selection of households was nevertheless set up to 

 include any household which might have been omitted in the prelisting. 

 This was done through use of the "half-open interval." Under this 

 procedure, the assignment of a household for interviewing actually 

 means the assignment of that household plus any other found to exist 

 between it and the next one on the list. In this way, the projections 

 reflect both the correction of any prelisting errors and changes in 

 the cluster since the prelisting took place. 



301 



