Questionnaire Developer (303388)*Madelon Cremers, Statistics Netherlands
Keywords: mode, design, omnimode, practical
The purpose of my presentation is to share practical implications of survey design. At Statistics Netherlands, designing and programming a computer-assisted questionnaire are two separate tasks, performed by two different teams. Survey designers design a questionnaire on paper, using Word for the text and Visio for the routing. Once the design is finalized, it is handed over to the survey builders who program the questionnaire in Blaise. Last, the designers test the computer survey by using test protocols, based on the on-paper design.
To respond faster to the market's demands, we thought about ways to optimize our survey process. One of the improvements we introduced was "omnimode design." In the past, StatNed always made two separate designs: one for the CATI/CAPI mode and one for the CAWI mode. Making two designs was necessary because of the many mode differences, e.g., the use of interviewer rules and cues, tables and nonresponse categories. Not only did this double our work, making two designs was also error prone, causing differences between the modes. By making a couple of changes in the interviewer instructions/cues and our design method, we are now able to make one design for all three modes: "omnimode." This not only shortened the length of our designing process, it also led to less errors and more satisfied interviewers. For now, only our on-paper design is truly "omnimode." Due to technical restrictions, it is still necessary to program two questionnaires (though programming does takes less time now than in the past). At the moment, StatNed works on a new IT and logistic system (Phoenix). In the future, we therefore hope to program only one questionnaire that incorporates all three modes.