12. Systemic Risk Emerging from Interconnections: The Case of Financial Systems

Sebastian Poledna
Elena Rovenskaya
Ulf Dieckmann
Stefan Hochrainer-Stigler
Igor Linkov

Systemic risk is the potential for a threat or hazard to propagate disruptions or losses to multiple nested or otherwise connected parts of a complex system. Systems prone to systemic risks are highly interconnected and intertwined with one another. Such interconnections contribute to complex causal structures and dynamic evolutions – typically nonlinear in their cause-effect relationships, often stochastic in their effect structure, and potentially global in their reach (in the sense that they are not confined within borders; International Risk Governance Center, IRGC, 2018). Systemic risks overwhelmingly do not follow normal risk distributions, but tend to be fat-tailed, i.e. there is a high likelihood of catastrophic events once contagion starts to unfold.

Systemic risk occurs in a wide variety of natural and human-made systems. It is the risk that a large part of the system ceases to function and collapses with potentially dramatic consequences for the system and its constituent parts. One of the most prominent examples of systemic risk today occurs in financial networks. Systemic risk in financial systems implies that a significant fraction of the financial system can no longer perform its function as a credit provider and collapses. The Great Recession started from the failure of a financial institution and propagated through the financial system, reaching also the real economy. In a broader sense, systemic risk also includes the risk of system-wide shocks that affect many financial institutions or markets at the same time.

Systemic risks in financial markets generally emerge through two mechanisms, either through synchronisation of the behaviour of agents (e.g. through fire sales, margin calls, or herding) or through interconnectedness of agents. The former can be measured by a potential capital shortfall over periods of synchronised behaviour, during which many agents are simultaneously distressed. The latter is a consequence of the network nature of financial claims and liabilities. Systemic risk is potentially extremely harmful because of the possibility of cascading failures, meaning that the default of one financial agent may trigger defaults of others. Secondary defaults might cause avalanches of defaults percolating throughout the entire network and can potentially wipe out the financial system via a deleveraging cascade. The fear of cascading failure is generally believed to be the reason why financial institutions under distress are often bailed out at tremendous public cost.

Financial systemic risk must not be confused with the default (single) risk of nodes or links in a networked system. The risk that financial agents primarily take into account is the so-called “credit default risk,” i.e. the risk that obligations, such as loans, are not paid at the agreed time, or not at all. This risk affects the lender immediately, but does not necessarily have systemic relevance. There exists an extensive literature on the understanding, regulating, and modelling of credit default risks. Current regulations of the financial system are almost exclusively focused on this type of risk. Credit default risks exist between two parties once they engage in a financial transaction, and usually no network aspects are considered.

The inability to see and quantify financial systemic risk arising from interconnections poses concerns of considerable financial and economic losses to society, and the failure to manage such systemic risk has been proven to be extremely costly. Systemic risk has become a focus of recent academic research, not only because of its societal importance, but also because of the availability of high-precision data enabling its qualitative assessment, and because the financial system is human-made and can in principle be changed and engineered to improve it.

The financial crisis of 2007–2008 was triggered by the default of a single investment bank. The consequences of this default propagated through the financial system, bringing it to the brink of collapse. Because of close links between the financial system and the real economy, the financial crisis spread quickly and triggered a global economic downturn, the so-called Great Recession. The majority of losses were indirect, such as people losing homes or jobs, and for the majority of people, income levels have dropped substantially. Despite such impacts, the mechanisms of how a financial crisis may lead to an economic recession, and vice versa, are not yet adequately understood at a fundamental level.

These developments have spurred research on systemic risk and financial networks. The clarification of the structure, stability, and efficiency of financial networks has become a hot research topic over the past decade. It has been shown that the topology of financial networks can be associated with probabilities of systemic collapse. In particular, network centrality measures have been identified as appropriate for quantifying systemic risk.

Systemic risk and financial contagion are largely related to synchronised behaviour and correlated portfolios of financial institutions. In this context, several econometric measures of systemic risk have been proposed that focus (mainly) on statistics of losses, accompanied by a potential shortfall during periods of synchronised behaviour, during which many institutions are simultaneously distressed. In particular, four statistical measures have been proposed recently: conditional value-at-risk (CoVaR), systemic expected shortfall (SES), systemic risk indices (SRISK), and distressed insurance premium (DIP). CoVaR is defined as the value at risk (VaR) of the financial system, conditional on institutions being in distress. The contribution to systemic risk of an institution is the difference between CoVaR conditional on that institution being in distress, and CoVaR conditional on that institution being in its median state. SES measures the propensity to be undercapitalised, given that the system as a whole is undercapitalised. SES is related to leverage and the marginal expected shortfall (MES). SRISK is closely related to SES and as such is a function of the size of an institution, its degree of leverage, and its MES. DIP measures the price of insurance against systemic financial distress in the banking system and is closely related to SES.

As an alternative to statistical measurements of systemic risk, it is also possible to take interactions directly into account and measure systemic risk in financial networks. Until recently, this alternative has been practically ignored by the mainstream economic literature. Research on systemic risk and financial networks has progressed only through the availability of high-precision empirical network data providing information on interbank networks, financial flows, or overnight markets. Several recent studies examine the evolution of financial networks and the network formation process. Their findings indicate that during the Subprime Crisis, a structural break appeared only after the collapse of Lehman Brothers; otherwise, interbank networks remained stable during this crisis. Research suggests that network measures can potentially serve as early warning indicators for crises. Several network-based systemic risk measures have been proposed recently. All approaches are based on quantifying the systemic importance of a node (institution) within a financial network. It has been reported consistently across many studies that the most relevant types of network measures for quantifying the systemic risk of a financial institution are network centrality measures. A disadvantage of such centrality measures is that their value for a particular node has no clear interpretation as a measure of losses due to systemic risk. An alternative to centrality measures that solves this problem is the so-called “DebtRank,” a recursive method suggested by Battiston et al. (2012) to quantify the systemic importance of nodes in terms of the losses a node would contribute to the total loss in a crisis. IIASA researchers have used DebtRank in a variety of studies to quantify systemic risk and have generalised DebtRank for multi-layer networks (Poledna et al., 2015).

Generally, empirical data on financial networks is not publicly available and is typically collected and owned by central banks or other government agencies. Because of the confidential nature of financial transactions, these agencies are reluctant to allow researchers access to this data. As a result, research on financial networks has mainly focused on credit networks between financial institutions. However, financial systemic risk is not only the property of a single network, but usually is determined by multiplex (or multi-layer) networks resulting from institutions being connected through various types of qualitatively different links, representing different types of financial contracts. Specifically, the layers of a financial multiplex network consist of the borrowing-lending contracts (obligations, i.e., counterparty exposures, and implicit relationships, such as roll-over of overnight loans), insurance (derivative) contracts, collateral obligations, market impact of overlapping asset portfolios, and networks of cross-holdings (holding of securities or stocks of other banks). Research on multiplex financial networks has appeared only recently. In collaboration with researchers from the Banco de México, the Mexican Central Bank, IIASA researchers from the Advanced Systems Analysis (ASA) and Risk and Resilience (RISK) programme analysed a financial multi-layer network. This work is based on a unique dataset containing various types of daily exposures between the major Mexican financial intermediaries (banks) over the period 2004–2013 (although for this work, data from 2007-2013 was used). Data were collected and are owned by the Banco de México, and various aspects of the data have been extensively studied. By evaluating contributions to systemic risk from four layers – (unsecured) interbank credit, securities, foreign exchange, and derivative markets – of the national banking system of Mexico, IIASA researchers have shown that focusing on a single layer significantly underestimates the total systemic risk (Poledna et al., 2015). Figure 12.1 shows the multi-layer financial network of Mexico, and Figure 12.2 shows how systemic risk evolves over time.

Another network layer of systemic risk emerges through common asset holdings of financial institutions. Strongly overlapping portfolios lead to similar exposures that are caused by price movements of the underlying financial assets. Based on the knowledge of portfolio holdings of financial institutions, Pichler et al. (2018) and Poledna et al. (2018a) quantify the systemic risk of overlapping portfolios. In particular, Pichler et al. (2018) present an optimisation procedure that enables the minimisation of systemic risk in a given financial market by optimally rearranging overlapping portfolio networks under the constraint that the expected returns and risks of the individual portfolios are unchanged. The developed approach has been applied to the overlapping portfolio network of sovereign exposure between major European banks by using data from the European Banking Authority’s stress test of 2016. It has been shown that systemic-risk-efficient allocations are indeed feasible. In the case of sovereign exposure, systemic risk can be reduced by more than a factor of two without any detrimental effects for the individual banks. The reduction of systemic risk is achieved by a dramatic decrease of the probability of contagion.

Not only financial firms, but also non-financial firms, such as vehicle manufacturers or energy companies, contribute to systemic risk in financial systems, in the same way as financial institutions as banks do. Poledna et al. (2018b) are the first to study the systemic importance of non-financial firms to shed light on mechanisms of how a financial crisis may lead to an economic recession, and vice versa. This work analysed data on nearly all financial and non-financial firms in Austria that included 80% of firms’ debts to banks. The researchers reconstructed the financial network between 796 banks and 49,363 firms, effectively representing the Austrian national economy in 2008, which is the most comprehensive financial network ever analysed. The researchers identified a number of mid-sized firms, with assets worth less than 1 billion Euros that are systemically important in the Austrian economy. This was previously unknown. Overall, the paper found that in Austria, non-financial firms introduce more systemic risk than the financial sector – 55% compared with 45%, respectively. This finding speaks strongly in favour of introducing regulations, similar to the Basel III rules imposed on banks to reduce the financial systemic risk they generate, also for non-financial firms. The results of this work could be the basis of a new approach to bank stress testing exercises that takes into account feedback effects between the real economy (goods and services) and the financial economy. Currently, bank stress testing exercises only assess the impact of risk drivers on the solvency of banks and are typically conducted without considering feedback effects among banks or between banks and the real economy.

In current financial regulation, systemic risk is regulated indirectly through capital requirements and other restrictions on financial institutions. On the regulators’ side, only in response to the financial crisis of 2007-2008, broader attention is now directed to financial systemic risk. A consensus is emerging on the need for a new financial regulatory system including a potential redesign of the financial sector: new financial regulations should be designed to mitigate the systemic risk of the financial system as a whole.

In the regulatory framework of Basel III currently under discussion, the importance of networks is recognised. In an effort directed at the reduction of systemic risk, the Basel Committee on Banking Supervision (BCBS) recommends future financial regulation for systemically important financial institutions (SIFIs). The Basel III framework recognises SIFIs, and in particular global and domestic systemically important banks (G-SIBs and D-SIBs), and recommends increased capital requirements for them, the so-called “SIFI surcharges.” By doing so, institutions are expected to alter their market behaviour and to internalise contagion externalities. Instead of using quantitative models to measure systemic importance, the BCBS suggests an indicator-based approach that includes the size of banks, their interconnectedness, their substitutability, their global (cross-jurisdictional) activity, and their complexity. In Poledna et al. (2017), IIASA researchers from the ASA and RISK Programmes, in collaboration with a researcher from the University of Oxford, studied and compared the consequences of different options for the regulation of systemic risk with an agent-based model and showed that Basel III would not reduce systemic risk in a substantial way.

Linkov et al. (2018) and Larkin et al. (2015) characterise the various strategies that United States federal agencies, as well as multiple directorates and affiliate agencies within the OECD, respectively assess and discuss systemic risk and threat. They individually frame the regulation and discussion of resilience through a disciplinary lens, where threat and system resilience are categorised into infrastructural, social, informational domains. Both pieces find that resilience is framed differently, with US Federal agencies placing greater emphasis upon infrastructural risk and resilience Linkov et al. (2018) with OECD placing greater emphasis upon social and economic concerns and systemic threats (Larkin et al., 2015).

Unlike for management of credit risk, proposals for management of systemic risk appeared only recently. While credit risk is relatively well understood and can be mitigated through several methods and techniques, management of systemic risk requires an understanding of the system as a whole. While it is evident that financial institutions as lenders have strong incentives to mitigate credit risk, it is less clear in the case of systemic risk as it involves externalities. In general, financial institutions manage their risks but do not consider their impact on the system as a whole. Management of systemic risk is, therefore, foremost in the public interest and must require financial institutions to internalise costs of systemic risk or otherwise create an incentive to minimise risks that are borne by the public.

Several authors have, therefore advocated various taxation schemes to manage systemic risk, while others are in favour of regulation due to the inherent difficulties of measuring systemic risk. Taxation schemes and the related measures for systemic risk are typically based on the notion of the systemic importance of financial institutions that need to be subjected to a Pigouvian tax. The idea is that institutions internalise contagion externalities if they are “taxed” based on their systemic importance. Levied taxes are typically collected by a rescue fund that can be used for bailouts. In Poledna and Thurner (2016), IIASA researchers introduced the notion of the marginal systemic risk, i.e., the systemic risk increment of an individual financial transaction, that is, its contribution to the overall systemic risk. Knowing the marginal systemic risk of individual transactions opens the way to an entirely new approach to managing financial systemic risk by reshaping the topology of financial networks. To apply this new approach, the researchers proposed a tax on individual transactions between financial institutions based on the marginal systemic risk that each transaction adds to the system and showed that this policy could essentially eliminate the risk of future collapse of the financial system.

In Leduc et al. (2017), an alternative mechanism to mitigate systemic risk by using credit default swaps (CDS) is examined. Considering that a CDS has the effect of transferring the default risk from one bank to another, the researchers showed that a CDS market could be designed to rewire the network of interbank exposures in a way that makes it more resilient to insolvency cascades. In Leduc and Thurner (2017), the authors used an equilibrium concept inspired by the matching markets literature to prove that the systemic risk tax proposed by Poledna and Thurner (2016) allows the regulator to effectively rewire the equilibrium interbank network to make it more resilient to insolvency cascades without sacrificing transaction volume.

IIASA research has also been used to study the economic and financial ramifications of crisis resolution mechanisms (Klimek et al., 2015). The authors of this study used an agent-based model, finding that, for an economy characterised by low unemployment and high productivity, the optimal crisis resolution concerning financial stability and economic productivity is to close the distressed institution. For economies in recession with high unemployment, the bail-in tool whereby debt of a financial institution is written off or converted into equity without shifting the burden to taxpayers, provides the most efficient crisis resolution mechanism. Under no circumstances do taxpayer-funded bailout schemes outperform bail-ins with private sector involvement.

The IRGC’s Guidelines for the Governance of Systemic Risk offer a risk governance approach that tackles the dynamic nature of complex adaptive systems (IRGC, 2018). Complex adaptive systems are in constant flux, and transitions between regimes are natural processes. Traditional probabilistic risk assessment methodologies cannot be successfully applied to risks that arise in such systems and may even have counterintuitive and unintended consequences. Since a system can be hampered by factors that reside inside or outside of its functioning as a complex system, dealing with systemic risks requires a dual process of identifying both problems and their interactions. Such notions are consistent with OECD discussions on systemic risk management and governance, such as the need to address global shocks and cascading failures, strengthen resilience, and create capacity for improved agility. Specifically, the IRGC Guidelines recommend a seven-step approach that is intended to help organisations identify, analyse, manage, and communicate their susceptibility to systemic risks:

  1. 1. Explore the system in which the organisation operates; define the boundaries of the system and the organisation’s position in a dynamic environment.

  2. 2. Develop scenarios, considering ongoing and potential future transitions.

  3. 3. Determine goals and the level of tolerability for risk and uncertainty.

  4. 4. Co-develop management strategies to deal with each scenario and the systemic risks that affect or may affect the organisation, and to navigate the transition.

  5. 5. Address unanticipated barriers and sudden critical shifts that may come up during the process.

  6. 6. Decide, test, and implement strategies.

  7. 7. Monitor, learn, review, and adapt.

Similar to IRGC’s approach, Linkov and Trump’s The Science and Practice of Resilience (Linkov and Trump, 2019) characterises systemic risk as a property of an organisation’s or system’s resilience. They use a definition proposed by the National Academy of Sciences (NAS, 2015) that frames resilience as the ability of a system to plan and prepare for, absorb and withstand, recover from, and adapt to adverse events and disruptions. Such disruptions can be sudden one-off events (shocks) or slow and even nearly imperceptible impacts (stresses). Linkov and Trump argue that systemic risk can be managed only by first understanding the core interdependencies and resilience (or lack thereof) within the various nested dependencies and critical functions of a given system, and then crafting countermeasures or system redundancies/fail-safes to ensure that a disruption to any given critical function will not trigger a cascading system failure. Specifically, systemic capacity to overcome systemic threat is framed as a particular measure of recovery and adaptation (Linkov et al., 2018).

As demonstrated by the review presented above, IIASA and OECD develop very complementary approaches to financial systemic risk. IIASA has a strong capacity in quantitative methods to measure, model, and manage systemic risk of financial systems using network theory and agent-based modelling. OECD looks into how to operationalise the concept of resilience to systemic risk to equip policy makers with an effective and efficient resilience management framework. IIASA’s quantitative methods can inform and enhance OECD’s framework by making available simple and transparent systemic risk indicators that can be monitored in real real-time, as well as tools to test alternative policy interventions to reduce systemic risk. For example:

  • Currently, financial regulations focus primarily on credit default risk ignoring risks generated through interconnections among banks or between banks and the real economy. New financial regulations, so called macro prudential regulation, should be designed to mitigate the systemic risk of the financial system as a whole and must require financial institutions to internalise costs of systemic risk or otherwise create an incentive to minimise risks that are borne by the public.

  • Current macro prudential regulation focusses almost exclusively on the financial sector. IIASA’s research indicated that in Austria non-financial firms introduce more systemic risk than the financial sector. This finding speaks strongly in favour of introducing regulations also for non-financial firms.

  • Bank stress testing exercises typically only assess the impact of risk drivers on the solvency of banks and are typically conducted without considering feedback effects among banks or between banks and the real economy. New approaches to bank stress testing should take feedback effects among banks and between banks and the real economy into account.

Approaches and models developed to deal with financial systemic risk may also have the potential to be useful to deal with systemic risk in other networked systems, for example, supply chains. Agent-based modelling framework developed by IIASA can be used to evaluate systemic economic consequences and indirect effects of natural disasters, whose frequency and severity are expected to increase due to climate change putting at risk economic growth and citizens’ well-being.


Battiston S. et al. (2012). “DebtRank: Too central to fail? Financial networks, the FED and systemic risk” Scientific Reports. Vol. 2(541)

IRGC (2018). “Guidelines for the Governance of Systemic Risks”, Lausanne: EPFL International Risk Governance Center (IRGC).

Klimek, P., S. Poledna, J.D. Farmer, & S. Thurner. (2015) “To bail-out or to bail-in? Answers from an agent-based model”, Journal of Economic Dynamics and Control, Vol. 50, pp.144–154i: https://doi.org/10.1016/j.jedc.2014.08.020

Larkin, S. Et al. (2015). “Benchmarking agency and organisational practices in resilience decision making”, Environment Systems and Decisions, Vol. 35/2, pp. 185-195

Leduc, Poledna and Thurner (2016), “Elimination of systemic risk in financial networks by means of a systemic risk transaction tax”, Quantitative finance, pp 1–15, https://doi.org/10.1080/14697688.2016.1156146

Leduc, Poledna, and Thurner. (2017) “Systemic risk management in financial networks with credit default swaps”, Journal of Network Theory in Finance, Vol. 3/3, pp. 19–39, https://doi.org/10.21314/JNTF.2017.034

Leduc and Thurner. (2017) “Incentivising resilience in financial networks”, Journal of Economic Dynamics and Control. Vol. 82, pp. 44-66, https://doi.org/10.1016/j.jedc.2017.05.010

Linkov, I., et al. (2018). “Resilience at OECD: Current State and Future Directions”, IEEE Engineering Management Review, Vol. 46/4, pp. 128-135.

Linkov, I., & B. Trump (2019). “The science and practice of resilience”, Springer.

Linkov, I., B. Trump, & J. Keisler (2018). “Risk and resilience must be independently managed”, Nature, Vol. 555/7694.

National Academy of Sciences (NAS). (2012). “Disaster Resilience: A National Imperative”, Washington, DC: The National Academies Press.

Pichler, A., S. Poledna, & S. Thurner. (2018). ”Minimisation of systemic risk as an optimal network reorganisation problem - the case of overlapping portfolio networks in the European government bond market”. https://arxiv.org/pdf/1801.10515.pdf

Poledna, S., S. Martínez-Jaramillo, F. Caccioli, & S. Thurner. “Quantification of systemic risk from overlapping portfolios in Mexico”, https://arxiv.org/pdf/1802.00311.pdf in preparation.

Poledna, S., A. Hinteregger, & S. Thurner. (2018) “Identifying systemically important companies by using the credit network of an entire nation”, Entropy, Vol. 20/10, pp. 792, https://doi.org/10.3390/e20100792

Poledna, S., Bochmann, O., & S. Thurner. (2017) “Basel III capital surcharges for G-SIBs are far less effective in managing systemic risk in comparison to network-based, systemic risk-dependent financial transaction taxes”, Journal of Economic Dynamics and Control, Vol. 77, pp. 230–246: https://doi.org/10.1016/j.jedc.2017.02.004

Poledna, S. et al. (2015) “The multi-layer network nature of systemic risk and its implications for the costs of financial crises”, Journal of Financial Stability, Vol. 20, pp. 70–81, https://doi.org/10.1016/j.jfs.2015.08.001

Battiston, S. et al. (2012) “DebtRank: Too central to fail? Financial networks, the FED and systemic risk”. Scientific reports, Vol.2/541, https://doi.org/10.1038/srep00541

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.


The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at http://www.oecd.org/termsandconditions.