A Guide to IMF Stress Testing

Chapter 28. Introduction to the Macro-Financial Approach to Stress Testing

Li Ong
Published Date:
December 2014
  • ShareShare
Show Summary Details
Andrea M. Maechler

During the global financial crisis, the world witnessed the near collapse of major financial institutions that many had believed to be impervious to the sort of panic that brought the global financial system to its knees and led to a worldwide recession. A key lesson from these events is that a purely micro-prudential approach to financial regulation and supervision, focused on the conditions of individual financial institutions and markets, could fail to detect important systemic or cross-cutting risks (Bernanke, 2011). Since then, there has been a generalized push globally to make the financial sector as a whole more robust by taking a broader macro-financial approach to oversight. The key feature of such an approach is to recognize that financial stability depends not only on the safety and soundness of individual financial institutions but also on how they interact with each other and with the real economy.

A macro-financial approach to financial oversight requires a holistic understanding of the sources of systemic risk. This includes monitoring the inherently procyclical nature of risk buildups (e.g., through leverage and rising asset prices), both in significantly important financial institutions and within the financial system more generally.1 It also means assessing risk correlations among individual market participants and understanding how they may be interlinked with other market participants. Finally, there is a need to explicitly incorporate macro-financial linkages in countries’ financial oversight frameworks, something that was poorly done before the global financial crisis (Bernanke, 2011).

In this regard, macro-financial stress tests have proven particularly useful in overcoming key shortcomings in traditional financial supervision. They have provided a consistent framework for assessing firms’ common exposures to a set of risks. Moreover, by linking key projected macroeconomic and financial variables to firms’ performance, they have also helped to improve the understanding of how shocks are transmitted through the financial system to the real economy and to provide a forward-looking assessment of financial stability.

Although not new, macro-financial stress tests have gained prominence over the last few years. Macro-financial stress tests have been an integral part of the IMF’s Financial Sector Assessment Program (FSAP) since the late 1990s, although they typically were not integrated into countries’ regular surveillance framework (Moretti, Stolz, and Swinburne, 2008). One noteworthy exception is the Bank of England, which has been implementing macro-financial stress tests since the early 2000s (Bunn, Cunningham, and Drehmann, 2005). In 2009, the U.S. authorities applied macro-financial stress tests under the Supervisory Capital Assessment Program (SCAP) to assess how much capital their largest bank holding companies needed to absorb potential losses and meet the needs of creditworthy borrowers under a more adverse scenario (Board of Governors of the Federal Reserve System, 2009a, 2011). In Europe, macro-financial stress tests were used to assess banks’ capital adequacy in the context of the sovereign debt crisis. At the IMF, they were used to conduct a regular capital shortfall exercise that would feed into their global financial stability assessment (IMF, 2008a, 2008b, 2009a, 2009b, 2010a). A stylized macro-financial stress test framework is presented in Figure 28.1 (see Bunn, Cunningham, and Drehmann, 2005; Jenkinson, 2008; and Constâncio, 2011, for discussions on macro-financial stress testing frameworks).

Figure 28.1A Stylized Macro-Financial Stress Test Framework

Source: Author.

There are two stages to the process:

  • In the first stage, an initial shock needs to be translated into a macro-financial scenario that maps the projection of key macroeconomic and financial variables (e.g., GDP, potential output, unemployment, house prices, bond yields). For example, market fears about an unsustainable fiscal position could result in higher sovereign yields, which could lead to a dampening in aggregate demand, higher unemployment, and lower potential output. This type of translation can be done using a variety of tools, ranging from large-scale in-house macroeconomic models (e.g., structural macroeconomic models, dynamic stochastic general equilibrium models) to smaller partial equilibrium time series models and vector autoregressive-driven impulse response functions.

  • In the second stage, the macro-financial scenario must be translated into firms’ financial conditions (e.g., firm losses and earnings potential). In an accounting world, this is done by mapping the shock onto firms’ balance sheet exposures. In a risk-based world, part of the adjustment takes place through changes in financial firms’ risk profiles (e.g., probability of default, loss given default, credit ratings). Market price–based risk approaches, including some of those presented in Part II of this volume, have the advantage of computing the distribution of possible outcomes, rather than single point estimates.

The strength of macro-financial stress tests is to produce a consistent set of impact assessments, although each bank will show a different vulnerability depending on its own risk exposures:

  • Credit risk models are the most developed. They can be based on accounting data or on more detailed portfolio-based or market-based data. Models linking nonperforming loans to macroeconomic variables are presented in Vazquez, Tabak, and Souto (Chapter 29) and Wezel, Canta, and Luy (Chapter 30). Sevogiano and Padilla (Chapter 31) present methodologies that incorporate the effects of macroeconomic shocks into credit risk, providing robust estimators when only short time series of loans exist and when only partial information about borrowers is available.

  • Market risk models, on the other hand, are still largely in their infancy. Different approaches have been used to capture market risk. The IMF’s U.S. FSAP capital shortfall exercise (IMF, 2010b), for example, focused on modeling the marked-to-market repricing of structured products (e.g., asset-backed securities, including mortgage-backed securities, credit default obligations). Among the high-profile stress testing exercises, the European Banking Authority stress test applied a fixed haircut to specific sovereign paper, while the U.S. SCAP exercise conducted a separate test to stress banks’ trading book.

Another challenge is to model the amplification of systemic risk through network effects. A severe strain at, or even default in, a systemically important financial institution, for example, is likely to affect the soundness of similar firms. This requires accounting for the inherently nonlinear nature of financial instability, where risk correlations can rise sharply in tail-risk scenarios. Promising work has been done in the area, including the approaches presented by Segoviano and Goodhart in Chapter 32 and by Jobst and Gray earlier in Chapter 26, where the entire nonparametric time-varying dependence structure among financial institutions is derived. Nonetheless, much remains to be done to fully integrate high-frequency, market-based approaches into the more traditional and slow-moving macroeconomic models.

Most macro-financial stress tests focus on the first-round effects, namely, on financial firms’ ability to absorb an adverse shock and provide credit to the economy. The impacts of severe macro shocks typically are modeled as if they occur on a one-off basis, leading to an underestimation of the possible impacts. The macro-financial shock is then mapped onto firms’ capital needs and credit growth potential. One example is the forward-looking balance sheet approach, such as the one used in the U.S. SCAP exercise and presented in a simplified fashion by Keim and Maechler in Chapter 33, which models the impact of a projected path for key macroeconomic and financial variables on banks’ capital buffers and other balance sheet positions, including their ability to provide credit. The crisis also has highlighted the challenge of aggregating risk factors across individual institutions and modeling firms’ profitability (see Board of Governors of the Federal Reserve System, 2009b), with stress testing revenues—especially for stressed conditions—largely seen to be a “black box” (Schuermann, 2012).

Thus, one of the biggest challenges for stress testing is to adequately model the systemic impact of a particular shock, taking into account the full cycle of macroeconomic and financial interactions and feedback effects. Individual bank responses to a given shock can lead to a broader market response, with further feedback effects to banks’ exposures. For example, a slowdown in issuance or a pullback of credit by banks following an initial economic or market shock could further affect economic activity (second-round effect), which would have implications for the banks’ ability to additionally absorb the consequences of further deterioration in the macroeconomic environment (third-round effect) and so forth. This requires macroeconomic models with richer real-financial linkages, something that largely was lacking prior to the global financial crisis (Roger and Vlcek, 2011). Tieman and Maechler (Chapter 34) take a step toward closing this gap in the literature by estimating the short-run feedback effects from financial sector risk to the real economy through the credit channel.


See, for example, Adrian and Shin (2008) on how leveraged investment can fuel an asset and/or credit bubble.

    Other Resources Citing This Publication