The Global Crisis and its aftermath led to a greater use of stress tests and to the establishment of macroprudential policy as a new policy area, with the objective being to identify and limit systemic risk. Early identification of risks, supported by thorough surveillance and early warning models to detect potential sources of systemic risk, is an essential first step in the macroprudential policy-setting process. Macroprudential is thus a pre-emptive policy. Furthermore, it must also be countercyclical: macroprudential policy should strive to smoothen the financial cycle by avoiding excessive risk with respect to the position of a specific economy along the cycle. Efforts have been made to develop analytical frameworks to support this new policy function.
Today, the ECB has published an eBook (Dees et al. 2017)1 which presents analytical tools that support design and calibration of macroprudential policy. The top-down models used to support EU-wide stress-testing exercises (primarily microprudential solvency exercises) have been elaborated to form the Stress Test Analytics for Macroprudential Purposes in the euro area (STAMP€), offering a suite of analytical tools which are part of the stress-testing framework developed by ECB staff over the past few years.
Financial sector stress tests have proved to be an important tool for assessing the robustness of the financial system and gauging risks arising at the system-wide level from a macroprudential perspective. In 2013, the ECB published an occasional paper describing the framework and its various modules for conducting stress tests (Henry and Kok 2013). These had been used since 2009 to support the EU-wide stress test by the Committee of European Banking Supervisors and later by the European Banking Authority (EBA). Since 2013, new modules and tools have been developed. These tools go well beyond the requirements of the traditional solvency stress tests applied to banks. They include a broader set of institutions than just banks, an analysis of the financial cycle, as well as an assessment of systemic risk levels associated with the economic and financial shocks considered in adverse scenarios.
Systemic risk measurement
In the ECB's Financial Stability Review, systemic risk is defined as “the risk that financial instability significantly impairs the provision of necessary financial products and services by the financial system to a point where economic growth and welfare may be materially affected”.2 At the heart of this definition is the notion that the materialisation of systemic risk imposes significant costs on the real economy.
The literature has identified three broad sources of systemic risk: (i) macroeconomic shocks that are significant enough to cause distress in the financial system, (ii) the unwinding of imbalances in the financial system generated by excessive leverage, and (iii) contagion risk, created by increasing interconnectedness and herd behaviour.
Several indicators to measure systemic risk have been proposed since 2008. The first type of indicators has a ‘micro-level’ dimension, i.e. it calculates the contribution of significant institutions individually to systemic risk. The Marginal Expected Shortfall (MES), Conditional Value at Risk (CoVar), CoRisk or Conditional Tail Risk (CRT), for example, fall into this category.3 Taken in isolation, they do not help to predict future levels of systemic risk, as they tend to use contemporaneous market prices and do not consider the system as a whole.
Composite indicators such as the ECB’s Composite Indicator of Systemic Stress (CISS) (Holló et al. 2012), represent a second type of measure. This comprises five aggregate market segments accounted for by a range of variables and time-varying rank correlations between them. Another example is CATFIN, a value-at-risk and expected shortfall measure at the system-wide level, calculated with non-normal distributions with fat tails, showing the predictive capacity of financial volatility regarding real economic downturns (Allen et al. 2012).
A third type of approach complements these efforts and relates to the concept of the financial cycle, in contrast to that of the economic or business cycle. In fact, a country’s positioning in the financial cycle – with respect to an historical benchmark – can be seen as a systemic risk indicator and can be used to predict overall levels of risk in the system. In that context, a first step was taken in a recent ECB working paper (Hiebert et al. 2016) building on and extending work done at the Bank for International Settlements (BIS) on the financial cycle (Borio et al. 2012). It shows how credit and asset prices share cyclical similarities, captured in a synthetic financial cycle index that outperforms credit-to-GDP gap measures in predicting systemic banking crises, on a horizon of up to three years.
Another complementary approach for assessing overall levels of systemic risk in the system could result from adding a macroprudential perspective to the analytical tools used in traditional bank solvency stress tests. In 2015, I elaborated on this whole concept, which seems to me a natural development (Constâncio 2015). An initial application of this approach in the wake of last year’s EBA stress tests was described in the ECB Macroprudential Bulletin of November 2016 (ECB 2016).
Limitations of traditional stress test methodologies
One important lesson of the Global Crisis was that the need to go beyond the micro-supervision goal of ensuring the robustness of individual financial institutions, particularly banks, was recognised. We learnt that the system can collapse even if, individually, institution by institution both solvency and liquidity positions seem quite safe. The degree of interconnectedness within the system, contagion through herd behaviour, and the sudden vanishing of inside liquidity within the financial system are realities that justify system-wide surveillance as well as the new macroprudential policy area. The ultimate objective of macroprudential policy is to prevent and mitigate systemic risk, which includes strengthening the financial system and smoothening the financial cycle, in order to preserve the effective provision of financial services to the real economy (Constâncio 2016).
In the aftermath of the crisis, microprudential solvency stress tests were promptly used to assess the capital needs of individual banks. However, this tool soon showed significant limitations in the face of macroprudential policy concerns.
Among these limitations is the static balance-sheet approach, which is not well suited to stress-testing exercises that run for a horizon of three years. This may render the tests unduly conservative if the macro scenario is too severe. No bank reaction is considered, whereas in practice, banks react to adverse conditions by deleveraging, making straight capital increases or working out non-performing loans. Each action of this kind would in turn have different macrofinancial consequences that would affect the economic environment.
Another weak feature is that the adverse scenario shocks are treated as exogenous to the financial sector and that feedback loops between credit institutions and the economy as a whole are ignored. Macroprudential stress tests should provide, in association with the adverse scenario, indicators to gauge the potential level of systemic risk related to each country’s position in the financial cycle. Higher capital requirements imposed on individual institutions may either not be enough to safeguard financial stability or, in different circumstances, may aggravate the overall financial stability conditions, requiring easing or release of macroprudential measures.
In the same vein, traditional stress tests do not include any interaction between banks and other specific sectors of the economy, whether households and corporates or other non-bank financial institutions, particularly asset managers and investment funds of all types.
Furthermore, no complete liquidity assessment is integrated in the microprudential solvency stress tests. This omission should be addressed, given the strong two-way interaction between liquidity and solvency strains brought to the fore by the global financial crisis.
These weaknesses can only be tackled in a true top-down macroprudential stress test framework, centrally conducted.
Macroprudential stress tests
STAMP€ is a relevant step towards providing an analytical framework for macroprudential stress tests. Beyond the macroprudential dimension of stress testing, the STAMP€ eBook presents top-down models that support the EU-wide stress-testing exercises – primarily microprudential solvency exercises – that are part of the overall framework (see Chapters 1 to 8). Regarding the macroprudential extension of stress testing, and corresponding to the shortcomings I mentioned before, the approach comprises five main domains to which analytical work presented in this e-book contributes.
- First, the dynamic dimension.
Macroprudential stress tests should encompass a dynamic approach that takes into account banks’ responses to the scenario. The tests need to account for realistic features of systemic stress, in particular banks’ behavioural reaction to the stress, as opposed to the static balance-sheet approach. Banks could react by deleveraging, raising capital or working out non-performing loans, for example. Typically, their reactions in the crisis caused the initial stress to escalate. This could be achieved by introducing a dynamic balance sheet that allows banks to re-optimise their portfolio according to the risk-return optimisation criterion (see Chapters 2 and 9), thereby departing from the traditional static approach.
- Second, the interaction with the real economy.
Macroprudential stress tests should take into account the two-way interaction between banks and the real economy as well as the related macro-feedback effects generated by banks’ balance sheet adjustments. To this end, a DSGE model calibrated for individual countries (see Chapter 10) and a Global Vector Autoregression (GVAR) model (Chapter 11) are being used, allowing for an assessment of the cross-border effects of deleveraging. Results tend to confirm the common wisdom that, in response to a negative shock to the leverage, banks tend to shed assets instead of raising capital, while keeping the leverage constant.
- Third, the interconnections between financial institutions.
Macroprudential modelling approaches need to account for interconnectedness among institutions and related contagion effects that can amplify the initial stress system-wide. Here, knock-on effects related to financial contagion and resulting from dynamic interactions between the financial economic agents are considered. Contagion via the interbank channel, already featuring in the ECB top-down stress-testing framework, is being refined (see Chapter 12). Stress-testing methodologies also need to include interaction with the financial sector, including the shadow banking sector, which continues to grow at a steady pace. These methodologies should help to reveal vulnerabilities in this sector and assess the potential for spillovers to the rest of the financial sector, most prominently due to fire sales. Agent-based models, allowing for endogenous asset price determination, can be used to account for such interactions (see Chapter 16).
- Fourth, the integration of system-wide liquidity assessment in the stress-testing framework.
There are two dimensions to this issue: one related with the interconnection between liquidity and solvency at the level of individual institutions; and the other, reflecting the consequences for overall liquidity associated with the interconnectedness and network effects within the financial system as a whole (see Chapter 14).
- Fifth and finally, macroprudential stress-testing methodologies need to account for interaction with non-financial sectors that are relevant for banks’ risk management.
A module presented in STAMP€ integrates the household sector of the economy in the stress-testing framework in order to properly account for vulnerabilities that may emerge from it (see Chapter 15). Using the data from the ECB Household Finance and Consumption Survey, a framework for stress testing balance sheets of households allows for the computing of the probability of default and loss given default for mortgage exposures directly at household sector level and links them to macroeconomic stress scenarios.
Such enhancements to the ECB top-down stress-testing framework also provide a tool for the impact assessment of macroprudential policy instruments. These policy instruments are designed and calibrated to address systemic risks. As such, assessing macroprudential measures in a consistent and holistic manner requires taking into account interactions within the banking sector as well as interactions with the financial and non-financial sectors and the real economy.
This publication represents the first step in a long journey, as reflected by the chapter on future extensions of the framework (see Chapter 16). It is a journey which goes well beyond banking, aiming at setting up a comprehensive stress-testing capacity for the financial sector as a whole. This ambitious endeavour requires the use of many different methods and models, from macroeconomic models to Bayesian Vector Autoregression (BVAR), to network contagion analysis or agent-based models. It requires putting together several analytical capabilities and the fostering of collaboration between many researchers and experts in different fields. To meet its responsibility for financial stability, the ECB is committed to continuing its work on developing an integrated framework appropriate to the assessment of the macroprudential policy stance. By publishing this eBook on the work in progress, we hope to engage the community of stakeholders in macroprudential issues, from academics to researchers in official institutions, in a fruitful dialogue and thereby improve the existing methodologies.
Allen, L, T G Bali and Y Tang (2012), “Does systemic risk in the financial sector predict future economic development?”, The Review of Financial Studies 25(1): 3000-3036.
Bisias, D, M Flood, A W Lo and S Valavanis (2012), “A Survey of Systemic Risk Analytics”, Office of Financial Research Working Paper No 1, January.
Borio, C, M Drehmann and K Tsatsaronis (2012), “Characterising the financial cycle: don’t lose sight of the medium term!”, BIS Working Paper No 380, June.
Constâncio V. (2015), “The role of stress testing in supervision and macroprudential policy”, keynote address at the London School of Economics Conference on “Stress Testing and Macroprudential Regulation: a Trans-Atlantic Assessment”, London 29 October. Also available in the VoxEU eBook, Stress Testing and Macroprudential Regulation: a Transatlantic Assessment.
Constâncio, V (2016), “Principles of macroprudential policy”, speech at the ECB-IMF Conference on Macroprudential Policy, Frankfurt, 26-27 April.
Dees, S, J Henry and R Martin (eds) (2017), STAMP€: Stress-Test Analytics for Macroprudential Purposes in the euro area, Frankfurt: ECB.
ECB (2009), Financial Stability Review, Frankfurt am Main.
ECB (2016), “Macroprudential effects of systemic bank stress”, Macroprudential Bulletin Issue 2, Chapter 1.
Henry, J and C Kok (eds) (2013), “A macro stress testing framework for assessing systemic risk in the banking sector”, ECB Occasional Paper No 152.
Hiebert, P, T A Peltonen and Y Schüler (2016), “Characterising the financial cycle: a multivariate and time-varying approach”, ECB Working Paper No 1846, September, and an updated presentation at the ECB-IMF Conference on Macroprudential Policy Frankfurt, 26-27 April 2016.
Holló, D, M Kremer and M Lo Duca (2012), “A composite indicator of systemic stress in the financial system”, ECB Working Paper No 1426.
 Dees, S, J Henry and R Martin (eds) (2017), STAMP€: Stress-Test Analytics for Macroprudential Purposes in the euro area, Frankfurt: ECB. Available at www.ecb.europa.eu/pub/pdf/other/stampe201702.en.pdf.
 See also Special Feature B in ECB (2009) for a discussion on the concept of systemic risk.
 For an overview, see Bisias et al. (2012).