Censoring Outdegree Compromises Inferences of Social Network Peer Effects and Autocorrelation

TL;DR

This paper investigates how incomplete edge data in social networks, due to censoring of outdegree, biases the estimation of peer effects and autocorrelation, revealing inflation or nullification of effects.

Contribution

It introduces an analysis of how censored outdegree impacts causal inference in social networks and evaluates methods to mitigate these biases.

Findings

Censoring outdegree can inflate effect estimates or bias them towards zero.

Heterogeneity in friend counts influences the effectiveness of post-hoc adjustments.

Different friend-naming constraints lead to varying biases in effect estimation.

Abstract

I examine the consequences of modelling contagious influence in a social network with incomplete edge information, namely in the situation where each individual may name a limited number of friends, so that extra outbound ties are censored. In particular, I consider a prototypical time series configuration where a property of the "ego" is affected in a causal fashion by the properties of their "alters" at a previous time point, both in the total number of alters as well as the deviation from a central value. This is considered with three potential methods for naming one's friends: a strict upper limit on the number of declarations, a flexible limit, and an instruction where a person names a prespecified fraction of their friends. I find that one of two effects is present in the estimation of these effects: either that the size of the effect is inflated in magnitude, or that the estimators instead are centered about zero rather than related to the true effect. The degree of heterogeneity in friend count is one of the major factors into whether such an analysis can be salvaged by post-hoc adjustments.