Selection a proper sample of households is not the end of the problem. It is still necessary to determine who in the household should be interviewed. The following are among the more common procedures at this step:
1) Talking to any adult who answered the telephone
2) Listing all the household adults and selecting the respondent from a table (roughly a random selection) and making up to three callbacks, if necessary, in an attempt to interview that person.
3) Same as #2 but substituting any other adult if the selected one is not at home.
4) Listing all household adults and selecting the youngest male or youngest female alternately.
5) Take first adult reached and then calling other numbers continuously until an adult of the opposite sex is reached.
6) Selecting respondents randomly (as in #2 above); but if sample tends to have more than half of one sex (usually too many women) select the other sex until the sex ratio is balanced.
One method for that requires the interviewer to ask two questions:
1) How many persons 18 years or older live in your household including yourself?
2) How many of these are women?
The interviewer then refers alternately to one of the four matrices shown in Table and selects the individual indicated. It is necessary to use all four matrices to balance males and females.
As shown in Table, if Matrix B were applicable and the household had three adults older than 18 and two women, the interviewer would select the one man for the sample. A procedure such as this will provide a representative sample, but it will take time to employ and will tend to create suspicion among those answering the telephone.
Response Rates: After a sample of telephone homes has been selected and a procedure established for selecting the individual within the household to be interviewed, it is still necessary to obtain the desired information. This is not as easy as it appears at first. Some numbers may not answer even after results from a telephone be survey in metropolitan Chicago where 5,346 b numbers were selected and as many as 10 calls were made, if necessary, to reach the household and the proper person within the household.
In general, one might estimate that to complete 100 interviews with randomly selected individuals in telephone homes, it would be necessary to have a list of 300 telephone numbers and to make 900 dialings.
If repeated efforts are not made to reach the households designated in the sample, the following groups will be overrepresented:
Final Disposition of Telephone Calls to sample of 5,346 Numbers in Metropolitan
Final Disposition Percent
Number not in service 33.0
Completed interviews 33.7
No answer after 10 calls 2.7
Business or other non-appropriate number 10.7
Foreign language required 2.5
No one older than 18 0.3
Selected respondent never reached 4.2
Refusal by household 6.3
Refusal by selected respondent 4.0
Break off during interview 1.8
Older than 64 age group
Low education group
Low income group
Respondents with home related occupations
Those outside large cities
Not at homes are somewhat lower in telephone surveys than in personal interviews at home because some people who will not answer the door to a stranger will answer the telephone. Of course, interviews made in shopping centers do not have this problem. Most investigators report lower refusal rates in telephone surveys than in personal interviews, but this can vary. Surveys on intimate and controversial topics tend to get larger refusal rates when done by telephone. A letter in advance significantly lowers refusal rates.