Network Connectivity and General Equilibrium Asset Prices

We will extend the framework paper presented in Branger et al. (2014), to compare networks, where links between the economic units are represented by mutually exciting jump processes, with networks, where connectivity is modeled via other, more conventional measures of comovement like joint negative jumps, regime-switching models or simple diffusive correlation. We will analyze how different ways of modeling tail risk and different measures to quantify network connectivity affect prices and returns in general equilibrium.

Another focus is the application of the model to empirically observe network topologies highlighted in the other projects. The cases we have analyzed so far in Branger et al. (2014) were motivated mainly theoretically, and we chose the networks such that the range of potential topologies is covered very broadly. But there are other interesting special cases. A typical banking network, e.g., consists of a small set of core banks which are highly interlinked, and a large set of periphery banks which are linked to only very few of the core banks, but to almost no other banks.

In order to compare different ways of modeling networks, we are currently working on a detailed empirical evaluation of the model in Branger et al. (2014). In particular, we analyze the quantitative effect of mutually exciting jump processes in state variables. These give rise to a notion of directedness of shocks, whereas other theoretical and empirical approaches can only motivate undirected measures of shock propagation (like, for instance, the popular eigenvector centrality). We test the impact of directed shocks as compared to undirected shocks using earnings data, aggregated at the industry level.