Using Randomization to Identify Social Influence in Mobile Networks

Identification of social influence in observational data is a difficult task. Endogeneity issues such as homophily, correlated unobservables and simultaneity raise challenges to the researchers interested in establishing causality and in consistently measuring its magnitude. In this paper we apply randomization techniques to identify social influence in a mobile network setting. Randomization methods consist in generating pseudo-samples of the original data by selectively permuting the values of some variables among observations, and estimating empirical distributions of a parameter of interest under the null hypothesis that such permutations are random. We show that randomization methods are a viable strategy to identify social influence in contexts where all adoption is observed and the date of adoption is available. Furthermore, we show that these methods provide a lower bound for the magnitude of the effect of peer influence. We use a comprehensive panel of data from a large European mobile carrier in one country. The data comprise Call Detailed Records for all the subscribers in this carrier for a period of 11 months. We also have information on pricing plans, adoption of products, promotions and handsets. We estimate the effect of peer influence in six of these promotions. We provide evidence for negative peer influence in their adoption. Peer influence reduces adoption for these promotions between 3% and 9%. Peer influence helps to share information about new promotions but also signals who has already adopted them and, in many cases, such as free calls, having neighbors who adopted the promotion is enough to benefit from it.

[1]  Mark S. Granovetter.  Threshold Models of Collective Behavior , 1978, American Journal of Sociology.

[2]  D. Watts,et al.  Influentials, Networks, and Public Opinion Formation , 2007 .

[3]  W. O. Kermack,et al.  A contribution to the mathematical theory of epidemics , 1927 .

[4]  Scott M. Smith,et al.  Computer Intensive Methods for Testing Hypotheses: An Introduction , 1989 .

[5]  Dylan Walker,et al.  Creating Social Contagion Through Viral Product Design: A Randomized Trial of Peer Influence in Networks , 2010, ICIS.

[6]  Cosma Rohilla Shalizi,et al.  Homophily and Contagion Are Generically Confounded in Observational Social Network Studies , 2010, Sociological methods & research.

[7]  Catherine E. Tucker.  Identifying Formal and Informal Influence in Technology Adoption with Network Externalities , 2008 .

[8]  Jennifer Neville,et al.  Randomization tests for distinguishing social influence and homophily effects , 2010, WWW '10.

[9]  S. T. Buckland,et al.  Computer-Intensive Methods for Testing Hypotheses. , 1990 .

[10]  Mark Wallace,et al.  The Effects of the Social Structure of Digital Networks on Viral Marketing Performance , 2008 .

[11]  Ravi Kumar,et al.  Influence and correlation in social networks , 2008, KDD.

[12]  E. Rogers,et al.  Diffusion of Innovations, 5th Edition , 2003 .

[13]  Frank M. Bass,et al.  A New Product Growth for Model Consumer Durables , 2004, Manag. Sci..

[14]  Arun Sundararajan,et al.  Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks , 2009, Proceedings of the National Academy of Sciences.

[15]  Ramayya Krishnan,et al.  Homophily or Influence ? An Empirical Analysis of Purchase within a Social Network , 2011 .

[16]  Jimeng Sun,et al.  A Survey of Models and Algorithms for Social Influence Analysis , 2011, Social Network Data Analytics.

[17]  Daniel G. Goldstein,et al.  The structure of online diffusion networks , 2012, EC '12.

[18]  Yogesh V. Joshi,et al.  New Product Diffusion with Influentials and Imitators , 2007 .