Summarized below are the approximate relations of 

 subgroup errors to errors of estimates from the entire sample. 



Type of Break 

 Region 



I ncome 

 City size 

 Household size 

 Age of nomemaker 

 Employment status 



Race 



Errors about double, except in West 

 where they are almost triple. 



Errors about 2-kk times as large. 



Errors about 2-3 times as large. 



Errors about 2-U times as large. 



Errors about 3-U times as large. 



Errors on full-time employed about 2 

 times, on part-time about 3g times, 

 on not employed about I5 times. 



Errors for Whites very close to same 

 as for all household 5; for ft on-White s 

 errors at least 3 times as large. 



For Protestants, errors about I4 times 

 as large, for Catholics about 2 times. 

 For Jewish, the errors are about 6 times 

 as large and hence are not useable in 

 many cases. 



Comparison of Sample Classification Data with Census Data 



Sample distributions of the classification data agree 

 reasonably well with the distributions as estimated from Bureau of 

 Census reports, wherever a basis for comparaoility exists. However, 

 it is unwise to place too much dependence on such comparisons, since 

 neither is the presence of close agreement with Census data a defin- 

 itive validation of the survey nor does the absence of such close 

 agreement in most cases necessarily invalidate the results. The 

 following are some comparisons of various classification data for 

 the sample to Census data: 



Religion 



310 



