Fuchs, Marek, Mick P. Couper, and S.E. Hansen. 2000. "Technology Effects: Do CAPI or PAPI Interviews Take Longer?" Journal of Official Statistics, 16(3), 273-286.
The transition from paper-based personal interviews to computer-assisted personal interviewing (CAPI) is already well underway. Much of the early research focused on operational issues and concerns about data quality differences between the methods. Attention is now being turned to more detailed assessments of specific features of the new data collection technology and its impact on the survey process.
This article deals with the question of relative administration of survey questions between paper and pencil and CAPI modes. Using data from a series of interviews (14 PAPI and 37 CAPI) using the National Health Interview Survey (NHIS) instrument conducted as part of usability testing of the instrument, a large number (over 2,200) of comparable items from the socio-demographic part of the instrument were subjected to detailed time and activity coding.
These data allow us to examine reasons for time differences across the modes. Where items are comparable in terms of design across modes, we find that CAPI takes slightly longer than PAPI, largely due to the speed of typing versus writing. However, most of the time differences found can be attributed to differences of design between paper and pencil and CAPI, rather than to the technology itself.
Since the early days of computer-assisted interviewing (CAI), the issue of whether computer-assisted interviews take longer than equivalent paper and pencil surveys has been discussed. Evidence in both directions can be found in the literature. To what extent does the time taken to administer survey questions depend on the technology (paper or computer) used? Or does the duration of an interview depend on the particular design features employed? In this article we explore this issue in detail, at the level of individual questions. Our goal is to understand whether and if so why the time taken to complete items on paper and on computer may differ.