Matrix design for social surveys

Convenor: Professor Christof Wolf
Affiliation: GESIS -- Leibniz-Institute for the Social Sciences

Session Details

General social surveys increasingly face problems of low response rates and increasing cost. In this situation alternatives to the traditional face-to-face mode are gaining interest. For these other modes, e.g. web surveys, the typical length of a general social survey is a challenge. Instead of 60 or more minutes one is probably restricted to 20 minutes. In this situation a matrix design for the questionnaire could be promising solution. In this design the original questionnaire is divided into modules and then only some of these are presented to groups of respondents. As a result we obtain a data matrix with many “holes” or missing data.
We welcome papers that address one or more of the following questions:
How to divide questions into modules? (How many modules? Thematic or random distribution of items? Should a core module be used or not? In how far does the specific mix of modules affect response behavior?)
How to distribute modules over respondents and how to implement this in the field? In particular if mixed mode approach is applied to data collection.
How must the resulting data be prepared to ease analysis? Should missing data be imputed? How can replicability be ensured? Do we need special methods to analyze the data?
Which problems does a matrix design pose for repeated cross-sectional studies concerning comparability over time? In the same way we can ask about the consequences for cross-national comparability?