Endogenous Network Formation: Experiments and Methods

Rong Rong

Advisor: Daniel E Houser, PhD, Department of Economics

Committee Members: Tyler Cowen, Carlos Ramirez, Robert Axtell

Truland Building, #400
April 26, 2013, 02:00 PM to 11:00 AM

Abstract:

My dissertation develops both substantive and methodological themes on the topic of social networks. Substantively, I conduct experimental studies based on the game theoretical models that describe network formation in various settings. Methodologically, I review the procedure of cluster analysis that could be used to discover the nature and the number of behavioral rules used by individuals in network environments.

Chapter 1: Growing Stars: A Laboratory Analysis of Network Formation

The acquisition and dispersion of information often occurs through social networks (Jackson, 2009).  Empirical and theoretical findings suggest that efficient information dispersion networks take the form of a star: small numbers of agents gather information for distribution to larger groups. Controlled randomized tests, however, have typically found little evidence of star network emergence. An exception is Goeree et al (2009), which reports reliable star network formation in an environment that includes ex ante heterogeneous agents. While heterogeneity may explain network formation in some environments, in others it may play a smaller role. Here we show that specific institutional environments promote star network formation in the presence of ex ante homogeneous agents. Especially effective institutions include investment limits and the “right-of-first-refusal,” both of which add stability to the decision environment. At the level of individual behavior, we find these institutions to encourage rational decision making and positive habit formation.

Chapter 2: Money or Friends: Social Identity and Truth Telling in Networks

Communication between departments within a firm may include deception. Theory suggests that small difference in monetary incentives explains why lying to outgroup members may be strategically optimal (Crawford and Sobel, 1982; Galeotti et al, 2012). In natural environments, however, social incentives also play an important role in determining the information people choose to share or to withhold. Unfortunately, little is known about how monetary and social incentives interact to determine truth-telling. We design a laboratory experiment to address this question. We found that absent social identity, players’ choices are mostly consistent with the theoretical predictions. Interestingly, the effect of identity is asymmetric: sharing the same identity does not promote truth-telling but holding different identities reduces truthfulness. We find that identity has an overall detrimental impact on truthfulness. These results have important implication for intra-organizational conflict management, suggesting that only by strengthening identity at the level of the organization can one create a positive impact on communication among different departments.

Chapter 3: Exploring Network Behavior using Cluster Analysis

Cluster analysis organizes a complicated data set into small number of groups based on patterns of similarity. It can be used to discover data structures without requiring strong ex ante assumptions about the properties of the data. Decision data from laboratory experiments are often generated by complex behavioral rules that can be difficult to specify a priori. These data may particularly benefit from clustering methods. This paper reviews key procedures and algorithms related to cluster analysis and discusses how to choose among clustering methods to analyze experimental data.