Abstract (first of two papers)
Norman M. Bradburn, Martin R. Frankel, Reginald P. Baker, and Michael R. Pergamit (5/91) "A Comparison of Computer-Assisted Personal Interviews (CAPI) with Paper-and-Pencil Interviews (PAPI) in the National Longitudinal Survey of Youth."
In discussions of mode effects, the survey methodology literature distinguishes three modes of data collection-face-to-face, telephone and self-administered. There is an extensive literature on possible effects of collecting data by each of these modes because they appear to differ in fundamental ways. What has been less noticed, however, is that there are variations within each of these methods regarding whether or not they are computer-assisted; that is, whether the questionnaire is represented in electronic or paper-and-pencil form. There is a paucity of literature on within-mode effects of using computers to assist in the data collection process.
We examined the differences for 139 variables between CAPI and PAPI cases in an experiment where assignments had been made randomly to mode of administration. Except for effects on interviewer errors that were programmed into the CAPI itself, in all of these comparisons we found only 4 differences that looked as if they might even approach statistical significance. This number is within the number that one might expect by chance when making multiple comparisons. There are a few differences, however, that deserve further study before rejecting them.
Abstract (second of two papers)
Norman M. Bradburn, Martin R. Frankel, E. Hunt, J. Ingels, A. Schoua-Glusberg, M. Wojcik, and Michael R. Pergamit (5/91) "A Comparison of Computer-Assisted Personal Interviews (CAPI) with Personal Interviews in the National Longitudinal Survey of Labor Market Behavior-Youth Cohort."
The purpose of this experiment was to assess the effect of conducting interviews in Round 12 of the NLS-Y by the Computer-Assisted Personal Interview (CAPI) method as compared with the traditional paper-and-pencil personal interview method. The experiment was conducted on one-half of the total sample and excluded respondents who had to be interviewed outside the United States and/or in Spanish. Interviewers were assigned cases in the same geographical region and, where possible, were matched with respondents for ethnicity. Assignment to the proper experimental or control group was done through random assignment of interviewers. Thus the experiment reflects actual field practices. The paper will report on the operational problems in conducting the experiment.
Last Modified Date: July 19, 2008