|
|
Authors conducted an analysis of contagion risk related to common exposures of banks to assets, covering a comprehensive set of asset categories. To this end, they defined a novel measure of contagion capturing a risk that a shock in a given market segment in a given country hits highly leveraged exposures. In other words, the losses implied by a unit size of a shock may erode a high portion of capital buffers.
Authors applied the constructed measure to a granular, confidential data set of the largest European banks reporting exposures, broken down by countries and sectors of the economy according to the European industry standard classification system (NACE). Moreover, they ran several experiments to assess contagion risk that influenced pricing of financial claims, based on the NEVA approach, accounting for pertinent network effects. In the simulations they benchmarked the results based on the observed network with contagion spreading in some stylized topologies of the financial system, reflecting either a full diversification of exposures or a concentration of exposures on the domestic market.
The second-round effects of adverse market conditions and related to banks’ interconnectedness are significant. Authors highlight this inherent feature of the financial system based on stylized simulations and confirm the importance of the network-based loss amplification channel by running their model under a plausible and consistent stress scenario used by European prudential authorities for their regular stress-test exercise.
Most importantly, they found that the diversified network of financial linkages provides a better cushion for a small-sized shock to the system than either the observed or a domestic configuration of the system. However, larger shocks imply a reversal of the relationship: the diversified system propagates the shocks much more severely than the domestic one. However, when using the relative equity loss as a measure of impact, they cannot observe that a more diversified structure is more stable for small shocks than a more domestically oriented structure. Interestingly, this second observation is at odds with the findings of Acemoglu et al. (2015), who worked in a more stylized set-up.