Abstract
This paper reports the use of verbal probes in cognitive interviews (CIs) undertaken to test the usefulness, validity and reliability of survey questions. Through examining the use of probes by three interviewers undertaking interviews as part the piloting of a cross-national crime survey, we examine which of the various types of probes used in CIs produce the most useful information. Other influences on interview quality are examined, including differences between interviewers and respondents themselves. The analyses rely on multi-level modelling and suggest that anticipated, emergent and conditional probes provide the most useful data. Furthermore, age, gender and educational levels appear to have no bearing on the quality of the data generated.
Acknowledgements
We would like to acknowledge the assistance of Ian Brunton-Smith for help with the multi-level modelling, and to the reviewers and editors, whose comments helped improve the clarity of our paper in places. Faults remain our own, of course.
Notes
All have been working on the FP7-funded Euro-JUSTIS project, designing new survey measures of citizens’ confidence in the criminal justice system
1. In retrospective CIs, all of the survey questions are delivered and answers elicited. The interviewer then asks the respondent a series of probes about these questions, taking each question or question set in turn.
2. However, herein we rely on data collected from just England and Finland. The process of coding in such detail and checking equivalence of coding decision across all four countries were felt to be too onerous to be conducted on all countries’ data. In the UK 25 and in Finland 24 CIs were completed.