Endogeneity Bias in Viral Marketing

Abstract

The last few years have witnessed a rapid increase in the use of social media by marketers. One of the main tenants of social marketing is that the structure of social networks helps drive marketing message over and beyond traditional methods such as advertising.Previous research has already established strong correlations between social contagion and brand equity in many settings. The goal of this paper is to investigate whether such correlations are the result of causal relationships, or other processes such as reactions to common, unobserved, shocks (hidden variable bias). This study is conducted with data from Facebook, the leading social network. We study the relationship between viral activities and the number of new users becoming fans of a brand in the context TV media.A baseline analysis, representative of current practices, is conducted and indeed we found a strong correlation between social contagion (viral activities) and brand equity (increase in fans). According to the baseline model, the margin impact is in the range of 12 to 18 new fans for every thousand users who get exposed to the brand through viral/social activities.Several models were constructed to identify the true causal relationship between social contagion and brand equity in this setting. Approaches including the use of instrumental variables, and structural modeling of the underlying endogeneity are employed to address the omitted variable bias, the simultaneity bias and other potential problems. We found that there is no significant causal relationship between viral activities and increase in fans. This result is robust across all modeling approaches. We believe this constitutes strong evidence that traditional methods may be over-estimating the impact of social contagion, and that it is important to address the aforementioned problems.