Simulating Fire-Sales in a Banking and Shadow Banking System

Abstract

We develop an agent based model of traditional banks and asset managers. Our aim is to investigate the channels of contagion of shocks to asset prices within and between the two financial sectors, including the effects of fire sales and their impact on financial institutions' balance sheets. We take a structural approach to the price formation mechanism as in Bluhm, Faia and Kranen (2014) and introduce a clearing mechanism with an endogenous formation of asset prices. Both types of institutions hold liquid and illiquid assets and are funded via equity and deposits. Traditional banks are interconnected in the money market via mutual interbank claims, where the rate of return is endogenously determined through a tâtonnement process. We show how in such a set-up an initial exogenous liquidity shock may lead to a fire-sale spiral. Banks, which are subject to capital and liquidity requirements, may be forced to sell an illiquid security, which impacts its, endogenously determined, market price. As the price of the security decreases, both agents update their equity and adjust their balance sheets by making decisions on whether to sell or buy the security. This endogenous process may trigger a cascade of sales leading to a fire-sale. We find that, first, mixed portfolios banks act as plague-spreader in a context of financial distress. Second, higher bank capital requirements may aggravate contagion since they may incentivise banks to hold similar assets, and choose mixed portfolios business model which is also characterized by lower levels of voluntary capital buffer. Third, asset managers absorb small liquidity shocks but they exacerbate contagion when liquid buffers are fully utilised.

Reverse Stress Testing: Scenario Design for Macroprudential Stress Tests

“…An agent-based model approach to model liquidity stress and the interaction with fire sales is explored in Calimani et al (2017).…”

“…• Characterization of families of worst-case scenarios: While the mechanism how fire sales can cause contagion from one geographic region to another is well understood from previous research (see, e.g., Calimani et al (2017) and Cont and Schaanning (2016), our findings highlight that for a distribution of realistic initial shocks, fire-sales contagion is maximized when losses spill over from one geographic region to another (as opposed to, e.g., fire sales amplifying losses within a given geographic region). In particular, applying our methodology to the EBA 2016 stress test data, our worst-case scenarios show that contagion occurs from a subgroup of banks located in Austria, Belgium, Ireland, Italy, Netherlands, or Sweden to a different subset of banks located in France, Germany, Italy, and the UK.…”

Reverse stress testing: Scenario design for macroprudential stress tests

“…Our main findings can be listed as follows: Systematic exploration of worst‐case scenarios : We develop an operational, algorithmic approach to search systematically for scenarios that maximize the contagion from distressed liquidations, under the assumption that agents will seek to minimize their losses when forced to deleverage. Characterization of families of worst‐case scenarios : While the mechanism how fire sales can cause contagion from one geographic region to another is well understood from previous research (see, e.g., Calimani et al. (2017) and Cont and Schaanning (2016), our findings highlight that for a distribution of realistic initial shocks, fire‐sales contagion is maximized when losses spill over from one geographic region to another (as opposed to, e.g., fire sales amplifying losses within a given geographic region). In particular, applying our methodology to the EBA 2016 stress test data, our worst‐case scenarios show that contagion occurs from a subgroup of banks located in Austria, Belgium, Ireland, Italy, Netherlands, or Sweden to a different subset of banks located in France, Germany, Italy, and the UK.…”

“…An agent‐based model approach to model liquidity stress and the interaction with fire sales is explored in Calimani et al. (2017).…”