Dependent Interviewing

Dependent interviewing is a technique used in some longitudinal research such as panel surveys, where the same individuals are reinterviewed at subsequent points in time. Depending on the survey design, interviews may take place quarterly, annually, biannually, or over some other specified period. Dependent interviewing can be defined as a process in which a question asked of a respondent within the current survey wave is informed by data reported by a respondent in a previous wave (Mathiowetz & McGonagle, 2000). For example, if, at one survey interview, a respondent reported being employed as a bus driver, dependent interviewing would use that response in the question asked about the respondent's job at the following interview. Instead of simply asking the respondent “What is your current job?” at both interview points, a dependent interviewing approach might ask, “Last year you told us you were working as a bus driver. Is that still [Page 251]the case?” The respondent can then either confirm the previous response or give a different answer if some change has occurred.

Longitudinal surveys are subject to the types of error normally found in survey data, including recall error, interviewer effects, and “noise” in the coding process. Longitudinal data, where continuous records of activities such as labor market histories or continuous income records are collected, suffer from an additional measurement problem known as the seam effect or the seam problem (Doyle, Martin, & Moore, 2000; Lemaitre, 1992). The seam effect occurs when an artificially high level of observed change in activity spells at the seam between two survey periods is reported by respondents. Respondents tend to report events that they have already reported in the previous interview and pull them into the current interview period, known as forward telescoping, with the result that events become bunched around the seam between the two survey periods. The use of dependent interviewing is designed to reduce these effects and produce more consistent data over time. Dependent interviewing has become more common with the advent of computer-assisted interviewing technologies for data collection, which have made it technically feasible to feed forward previously reported information from one interview into the next interview. Dependent interviewing has benefits in terms of data quality, reducing respondent and interviewer burden, as well as the cost of data collection and processing. Despite this, there are also concerns. Depending on how it is implemented, dependent interviewing could result in an underestimate of the true level of change within the population because it is easier for the respondent to simply agree with his or her previous report.