UCL Department of Economics


Job Market Paper

"An Econometric Model of Network Formation with an Application to Board Interlocks between Firms"


This paper provides a framework for studying identification in a network formation model. Network formation is modelled as a static game with complete information and pure strategy equilibrium. Links formed by players have directions. Payoffs depend on some players' characteristics partially observed by the researcher and on an externality, or spillover effect (hereafter SE), that goes beyond direct connections - i.e., player i's payoff from forming a link with player j monotonically depends on the number of additional players creating a link with j. This implies that parameters in players' payoffs are partially identified without further assumptions on equilibrium selection. The set of admissible parameter values (sharp identified set) is derived. Even if restrictions are added, conducting inference on the sharp identified set is prohibitively complex when there are four or more players. To attenuate the computational difficulties, the focus is on a larger set of parameter values (outer set) obtained by bounding the empirical probability of any network section being the unique equilibrium, and the probability of such a network section being a possible equilibrium, in a local game of the network formation game. The suggested outer set shows advantages over other outer sets in the literature (Tamer, 2003; Ciliberto and Tamer, 2009; Sheng, 2014), in terms of computational tractability and width. A 95% confidence region for the characterised outer set is computed using data on board interlocks between Italian firms. Results reveal that SE has a positive sign, i.e., firm i's payoff from forming a board interlock with rival j increases with the number of additional competing firms creating a board interlock with j. In view of the co-optation theory in corporate governance, this seems to support the idea according to which the higher the number of competitors with a director sitting on j's board, the stronger their capacity to influence j's decisions and align them with the group's interests.