Estimating Network Size with Aggregate Relational Data: Does Mode of Survey Administration Matter?

• In Event: New Advances in Social Network Analysis

Abstract

We present findings from an experiment that examines how mode of administration affects mean degree and accuracy of responses collected from an aggregate relational data (ARD) survey, which is a common approach to network size estimation. In this approach, people are asked to report on the number of people they know in populations of known size (“how many X’s do you know?”). Responses to these questions are scaled up to reflect the proportion of people one knows in a certain geographic area, which is used to calculate degree. Though such surveys were traditionally administered in person or over the telephone, they are increasingly collected on-line or through self-administered computer assisted systems. Given the high degree of cognitive burden associated with ARD questions, it is important to ask whether self-administered ARD surveys and face-to-face surveys yield data that are of similar quality. To answer this question, we constructed a two-condition laboratory experiment that randomly assigns college student participants to either a self-administered or a face-to-face ARD data collection process. We have collected data from 119 participants to date and are on pace to collect data from our target of 250 participants by the end of the semester. If we find that the self-administered ARD survey yields data that are as high or higher quality than data collected by interview, then it suggests new opportunities for widespread ARD data collection. If we find that data collected through the self-administered process underperforms, then it suggests that ARD researchers should invest time and resources in interviewer administered ARD questionnaires. In addition, if we find differences in degree estimates across modes, our findings may help to explain patterns in degree estimates in the existing ARD literature. Our findings will thus advance understanding of ARD methods and the implications of those methods for degree estimates

Authors

• Jody Clay-Warner, University of Georgia

• Derick Baum, Brown University

• Justine Eatenson Tinkler, University of Georgia