The Impact of Computer-Assisted Interviewing on Interview Length

Abstract

Computer-assisted personal interviewing (CAPI) offers many attractive benefits over paper and pencil interviewing. There is, however, mixed evidence on the impact of CAPI on interview length, an important survey outcome in the context of length limits imposed by survey budgets and concerns over respondent burden. In this paper, recent data from a large, nationally representative panel study is used to investigate CAPI’s impact on interview length. A key feature of our analysis is that, through use of both experimental and quasiexperimental evidence, we examine the roles played by specific factors which, while typically associated with CAPI, vary in their extent and nature from study to study. We find that effects very much depend on how CAPI is implemented: the hardware and software adopted, the extent and nature of the dependent data introduced, and even interviewer workloads, can all have large influences on the CAPI impact — a finding that helps explain the conflicting results from previous studies. Overall, we conclude that, absent dependent data, CAPI will almost certainly increase interview lengths. However, the potential reductions in interview length from dependent data are very large, such that even modest levels of dependent data can lead to net reductions in interview lengths.

Keywords: Computer-assisted methods, interview length decomposition, dependent data