Hagan and Collier Selection Method

The ideal method of selecting respondents within households is a probability method, but researchers seldom work under ideal conditions. They usually seek to improve within-unit coverage without adding to non-response. Probability selection methods, such as the Kish method, may enhance within-unit coverage but also may increase nonresponse because of the need for longer introductions, which are potentially intrusive and may alienate respondents. Longer surveys add to interviewer costs. Hagan and Collier developed a non-probability quota technique that they believed would secure respondent cooperation more readily because it asked no household composition questions. Their procedure was based on the Troldahl-Carter-Bryant (T-C-B) respondent selection method, which in turn was based on the Kish technique.

Interviewers are given four forms, which are randomly distributed. The result is an oversampling of male respondents (desirable because of higher non-contact-related nonresponse among men). Within the eligible age range for a given survey, Form A requests the youngest male in the household; Form B asks for the oldest male; Form C, the youngest female; and Form D, the oldest female. If no such person resides there, the interviewer is instructed to ask for the person in the household of the opposite sex and same age grouping. Forms A, B, and C are each used two times in seven, and Form D is used one time in seven.

Hagan and Collier reported favorable results in a sample compared with one that selected respondents by the T-C-B procedure. Demographic characteristics were similar, and the refusal rate at respondent selection was almost 5% less than the T-C-B method. Both methods have a small within-unit coverage bias because adults in households of more than two adults of the same sex whose ages are between the oldest and youngest adults have no chance of selection. Also, in three-adult households, one of the three adults would have the chance of being designated the respondent twice. Troldahl and Carter considered these violations of random sampling to be very small. Research using census data has shown that the bias caused by the Hagan-Collier method is very slight.

An example of Hagan-Collier question wording is May I please speak to the “youngest man”? Another example is For this survey, I need to speak with the youngest adult male in your household over the age of 17, if there is one. If there is none, the following question is asked: Then may I please speak with the youngest adult female? Wording should include the fact that the designated respondent is not the one who happens to be at home at the time but, instead, is the one who lives in the household. Interviewers need training in awareness that a woman in a one-person household fits as either the youngest woman or the oldest woman, that “youngest man” can apply to an elderly male, and that informants can be confused and think the interviewer is asking for an old man (or a young woman), among other things.