Li, Muxin (2016). Instrumental Variable Estimation of the Causal Eﬀect In Social Network. 12th Annual Indiana University Department of Economics’ Jordan River Economics Conference.
The recent development of internet markets and social networking has made available a great quantity of data regarding individual's social connections and their consumption habits. However, one important underlying prerequisite of this marketing strategy is that people's consumption behaviors are a ected by others in the same social network. Past researchers support this argument by illustrating the positive correlation between people's behaviors in the same network. However, those arguments are not strong enough to prove the causal relationship between individuals' consumption and the network because this positive correlation of consumption behaviors could come from other resources like homophily and common confounding factors. In this paper, we use instrumental variables to prove the causal effect of network towards individual's consumption behaviors. In addition, we also explore the specifc position in the network by which individual's consumption could be better predicted and hence help us determine the standard of choosing the best marketing targets in the social network. To check whether our conclusion and strategy could apply to different products, weuse both the consumption data of alcohol and cigarette consumption to test our hypothesis. We propose an econometric method for estimating two kinds of peer effects on consumption, neighbor peer effect and competitive peer effect. We make use of instrumental variables to make peers' consumption exogenous and prove the causal effect of peer's behavior on individual consumption.
12th Annual Indiana University Department of Economics’ Jordan River Economics Conference
City of Publication