Respondent Selection bias in the Hagen-Collier approach

Abstract

When more than one household member is eligible for a survey it is necessary to devise a method that will assure that every eligible member has an equal chance of being selected. The random selection of a respondent in a household would seem at first to be a very simple operation. In reality, it poses several challenges largely due to the need to ensure that the interviewer does not introduce bias, intentionally or otherwise. The most elaborate method for achieving this goal is known as the "Kish table" method which generates for each household a random number selection table which in turn indicates which respondent to choose depending on the size of the household, i.e., the number of eligible members (Kish, 1965). There is no argument that this method is guaranteed to ensure random selection. However, the disadvantages are, first, the screening form becomes more complicated; second, the application of the Kish tables requires more work of the interviewer; third, it is intrusive, since a complete household roster is required; and fourth, it lengthens the introduction to the interview which is the point during which most nonresponse takes place.