Econometric analysis of network formation models

This dissertation addresses topics in the econometrics of network formation models. Chapter 1 provides a review of the literature. Statistical models focus on the specification of the probability distribution of the network. Examples include models in which nodes are born sequentially and meet existing vertices according to random meetings and network-based meetings. Within this group of models, special attention is reserved to the milestone work by Jackson and Rogers (2007): after having discussed and replicated the main results of the paper, an extension of the original model is examined and fitted to a dataset of Google Plus users. Even if statistical models can reproduce relatively well the main characteristics of real networks, they usually lack of microfundation, essential for counterfactual analysis. The chapter hence moves to considering the econometrics of economic models of network formation, where agents form links in order to maximise a payoff function. Within this framework, Chapter 2 studies identification of the parameters governing agents’ preferences in a static game of network formation, where links represent asymmetric relations between players. After having shown existence of an equilibrium, partial identification arguments are provided without restrictions on equilibrium selection. The usual computational difficulties are attenuated by restricting the attention to some local games of the network formation game and giving up on sharpness. Chapter 3 applies the methodology developed in Chapter 2 to empirically investigate which preferences are behind firms’ decisions to appoint competitors’ directors as executives. Using data on Italian companies, it is found that a firm i prefers its executives sitting on the board of a rival j when executives of other competitors are hosted too, possibly because it enables i to engage with them in “cheap talk” communications, besides having the opportunity to learn about j’s decision making process