Detecting measurement bias in respondent reports of personal networks

Inaccuracy of sociometric reports poses a serious challenge to social network analysis. Nevertheless, researchers continue to draw potentially misleading conclusions from flawed data. We consider two particular types of systematic error in measurement of network size: individuals over/underreporting others (expansiveness bias), and individuals being over/underreported by others (attractiveness bias). We examine evidence of individual variation in these biases in one apparently typical sociometric dataset. We specifically suggest that variation in expansiveness bias may commonly distort findings concerning characteristics of individual networks (e.g. size, range, density), and properties of whole networks (e.g. inequality, transitivity, clustering, and blockmodels). We suggest methodological improvements and urge further research. © 2002 Published by Elsevier Science B.V.