Systemic risk is responsible for large social costs (5% of GDP in assistance to the financial sector, ECB estimates 2015). Correctly assessing systemic risk and identifying banks that are vulnerable and have at the same time a large impact on the economy is crucial to mitigate costs in the future.
As the 2007-08 financial crisis showed, standard stress-test methodologies severely underestimate risk. This is because banks form a complex, highly-interconnected network, in which risk is amplified through bilateral exposures.
This DebtRank dashboard represents the main visual output of a novel stress-test framework to monitor systemic risk in financial systems. The paper describing the framework can be found here (SSRN) or here (arXiv). A pdf version of the dashboard can be downloaded here.
We reconstructed the interbank lending network from analyzing the links between institutions’ balance sheets and developed a new stress-test in which external shocks are mechanically amplified through the network of bilateral exposures.
The stress-test is divided into three main rounds:
- First Round: a negative fixed shock of 1% impacts the external assets of all banks, determining an increase in the vulnerability of all institutions, according to their leverage to external investments,
- Second Round: the initial shock reverberates through the interbank lending network. The losses incurred in the first round increase the probability of default in borrowing institutions, making them less likely to repay their debts and thus affecting their lending counterparties in the interbank lending network,
- Third Round: banks attempt to restore their original leverage levels by selling external asset en masse (fire-sales).
The dashboard below is interactive and shows the results of this stress-test obtained by implementing the DebtRank algorithm on the network of the top 50 European banks by assets. Explore the disaggregated effects of the various rounds, as quantified by the global vulnerability of the banking system, i.e. the total relative equity loss of the network. It is clear from our results that not taking into account network effects results in a severe underestimation of systemic risk.
This stress-test framework allows us to study the vulnerability of individual institutions together with their impact on other banks, quantified as the total relative equity loss induced on the network by their default. The presence of high-vulnerability and high-impact institutions constitutes a typical systemic risk that is overlooked by standard stress-testing methods. Move the cursor over the years and explore how the impact and vulnerability of favorite bank changed through time, particularly during the financial crisis. Institutions occupying the top-right corner of the graph are said to be systemically important banks.
The flexibility of the stress-test framework allows us to implement various financial distress propagation algorithms. The Eisenberg & Noe model (together with its revised formulation, the Rogers-Veraart model) is one of the most popular mechanisms for distress propagation in financial networks. One of its core assumptions is that a bank propagates financial distress in the network only after it has defaulted, resulting in a severe underestimation of global vulnerability. As the following graph shows, the DebtRank algorithm, allowing distress propagation by both defaulted and non-defaulted institutions, captures more clearly the second round effects, providing a unique tool in monitoring the fragility of the banking system.
The following interactive dashboard illustrates the preliminary results of an incoming paper on climate stress-testing. The aim is to provide a systemic risk assessment of the exposure of the EU banking sector to climate policy shocks. By implementing an hypothetical shock profile on single constituent companies of specific energy sectors in the EU economy, we have been able to study the financial impact on their shareholders, among which are many EU banks.
Original publication date: 6 November 2015.
The SIMPOL Project is currently funded by the H2020 European grant DOLFINS (no. 640772) in the Global Systems Science area of the Future Emerging Technologies program.