I used to work in market ressearch (a long time ago) there was 2
schools of survey then: 1 population sample, ie. you reflect the
composition of the population on your survey sample:
eg pop: 10 millions 20 % executive, 40 % blue collar, 10 %
unemployed age groups 20 % below 15, 20 % 16-25 etc. on a sample
of 100 people you should have 20 executives, 40 bc, 10 unemployed and
their age should match the percentage of the population age.
 in the same proportion, the interviewers were given very strict
quota. The problems occurs when at the end of your survey you need to
interview a 15 year old woman executive, we allowed 5% margin of error
. The other system used usually for large surveys (5000 +) was called
caterpillar and you pick 1 house in 3, 1 door in 3, etc that system
never proved terribly good as the interviewers would avoid 'less
friendly looking houses' and thus flaw the system.-
Dominique [log in to unmask]
Ganapati Kumari, Pinmill, Ipswich, IP9 1JW, UK.