Influence Versus Selection: A Network Perspective on Opinion Leadership

Abstract

Over the past decades, research on opinion leaders has been based on an implicit assumption that the structure of social networks is stable and that only attitudes and behavior are subject to change in a diffusion process. The finding that social groups often display similar attitudes or behavior was therefore regarded as evidence of opinion leaders’ influence. However, network autocorrelation also can emerge due to social selection processes in which likeminded people establish new ties and cut dissonant ties. In fact, without controlling for social selection processes, one is likely to overestimate the power of influence processes. Stochastic actor-oriented modeling of dynamic social networks allows disentangling and quantifying these two processes. This reanalysis of a four-wave panel survey of adolescents’ conversation networks and their TV use is the first to do this on the level of specific TV programs. The results demonstrate that influence of opinion leaders may become insignificant if parameters for social selection and general patterns of program preferences are included in the analysis. Overall, this study challenges an overly simplistic view of opinion leadership and illustrates the power of longitudinal social network analysis for disentangling social influence and social selection processes.