Birds of a Feather Flock Together: Distinguishing Social Influence from Shared Personality Traits
(A2019-9959)

AUTHORS

Jeroen van den Ochtend, University of Zürich; Raluca Gui, University of Zürich; Markus Meierer, University of Zürich

KEYWORDS

Social Influence; Homophily; Copula Method

ABSTRACT

A prominent challenge faced in research on social influence is to prove that locally clustered customer behavior stems from social influence and is not caused by unobserved homophily, i.e. unobserved personality traits that connected customers have in common. The authors of this paper approach this problem as omitted variable bias and suggest an internal instrumental variable method in the form of the Copula approach to retrieve an unbiased estimate of social influence. The authors conduct a simulation study to compare the performance of the Copula approach to two alternative approaches to model latent homophily; the latent space approach and the spatial autoregression approach. The results provide evidence that the Copula approach outperforms the alternative approaches in reducing the bias of the social influence parameter estimates. To help scholars deciding which approach would be the most suitable for their research, the strengths and weaknesses of each approach are discussed.