Troldahl-Carter-Bryant Respondent Selection Method

Researchers often desire a probability method of selecting respondents within households after drawing a probability sample of households. Ideally, they wish to simultaneously maximize response rates (by [Page 909]gaining the respondent's cooperation) and within-unit coverage (by obtaining an accurate listing of all eligible persons in a household). However, they need to balance these two goals because no selection technique is perfect at accomplishing both. In the early history of telephone surveys, one procedure that was considered easy, quick, and likely to improve respondent cooperation was known as the Troldahl-Carter method, to which Barbara Bryant later suggested some modifications. Face-to-face surveys commonly used the Kish technique, which asks for a listing of all men and all women in the household, ranked by age.

Verling Troldahl and Roy Carter feared that informants (those who answer the phone) in a telephone survey would become suspicious of questions that try to obtain a listing of all eligible household members and refuse to participate. So they suggested a brief procedure, building on Kish's work but requiring only two questions: (1) How many persons (18 years or older) live in your household … counting yourself? and (2) How many of them are men? Once interviewers knew the basic household composition by sex and age, according to responses to the two questions, they requested the appropriate respondent after consulting one of four versions of simplified selection tables assigned to households randomly. Because this selection method specified that the respondent be the youngest or oldest male or female, it involved a small violation of random sampling and full coverage because some adults in households of three or more adults of the same sex had no opportunity to be chosen. The amount of bias depends on the proportion of persons in the population barred from the sample and the degree to which they differ from the respondents in the variables studied. In addition, in three-adult households, one of the adults would have the chance of selection into the sample twice. Troldahl and Carter thought these biases were minimal. (Kish's plan also contained a small violation of random sampling, although it is considered to be an almost pure probability method.) In subsequent testing of the Troldahl-Carter method, other modifications to it were suggested, such as changing the second question to How many of them are women?

Troldahl and Carter provided an example of an interviewer's selection sheet. The potential numbers of adults in the household were shown at the top (1, 2, 3, or 4 or more). Interviewers circled the correct number and drew a line down the column under that number to the number of men in the household, ranging from 0 to 4 or more. For example, if there were three adults in the household and two were men, one version of the tables showed that the column at which those items intersected designated the respondent as “youngest man.” The interviewers' instructions were to say, I have to ask some questions of the [PERSON SELECTED] in your household. If the informant was of the correct gender, the interviewer asked, Would that be you? If the desired respondent was of the opposite sex, the interviewer asked for the chosen person and implemented follow-up procedures if he or she was unavailable.

...