The Social Contagion Hypothesis: Comment on "Social Contagion Theory: Examining Dynamic Social Networks and Human Behavior"

TL;DR

This paper critically examines the statistical methods used in social contagion studies, highlighting challenges in establishing contagion effects from observational network data despite methodological efforts.

Contribution

It provides a critical analysis of existing methods for detecting social contagion, emphasizing limitations and sensitivities in observational data analysis.

Findings

Methods are sensitive to model specifications

Contagion effects are difficult to establish from observational data

Caution is needed in interpreting network-based contagion results

Abstract

I reflect on the statistical methods of the Christakis-Fowler studies on network-based contagion of traits by checking the sensitivity of these kinds of results to various alternate specifications and generative mechanisms. Despite the honest efforts of all involved, I remain pessimistic about establishing whether binary health outcomes or product adoptions are contagious if the evidence comes from simultaneously observed data.