| dc.description.abstract | This thesis examines contagion in networks in two settings: narratives infection in social networks and default contagion in nancial networks. In the rst chapter, we consider the choice of a politician of the timing of disclosure of a narrative, to maximise its adoption at a given deadline. Voters update their opinions following a variation of the average-based De Groot learning concept. With a certain probability a voter adopts an alternative narrative, and the politician faces a trade-o between early and late disclosure. We nd the optimal timing of disclosure for disconnected and connected voters in certain networks. Finally, we consider the impact of homophily on timing of disclosure and explore comparative statics. Chapter two presents a model of default in nancial networks where the decision by one agent on whether or not to default impacts the incentives of other agents to escape default. Agents' payo s are determined by the clearing mechanism introduced by the seminal contribution of Eisenberg and Noe (2001). We rst show the existence of a Nash equilibrium of this default game. Next, we develop an algorithm to nd all Nash equilibria that relies on the nancial network structure. Finally, we explore some policy implications to achieve e cient coordination. In the last chapter, we extend the model of exposure of nancial institutions through liabilities linkages introduced by Eisenberg and Noe (2001), and we explore a second channel of exposure through credit lines or promised payments. While the liabilities create direct risk of default, credit lines can cause the default of their potential receivers. We prove the existence of a unique payment equilibrium and establish a ctitious default algorithm that computes the payment equilibrium and the chain of defaults. Finally, we investigate approaches to mitigate risks through multilateral netting and central clearing and by targeting banks with cash injections. | en_US |
