« Previous | Main | Next »

June 26, 2009

Stress Testing: An extraordinary balancing act

The IMF held an unofficial discussion yesterday morning on what the financial crisis has taught us about stress tests. (The discussants were very clear—none of the views expressed at the meeting represent the official views of the IMF. The same is true of the summary here.) 

The conversations highlighted the delicate balance that such tests need to achieve. When creating the parameters for these tests, how do we "think the unthinkable", while still adopting reasonable parameters that provide credibility? How do we decide if a 25 percent exchange rate depreciation is enough, or if 50 percent is too much? How do we design a scenario that is exceptional, but plausible?

Take, for example, the most spectacular banking system collapse of the entire crisis: Iceland. A case study of Landsbanki Islands, the country's second largest bank, applied the following stress test criteria:

  1. 35% fall in domestic stock prices, 25% fall in foreign stocks;
  2. 20% loss in the value of non-performing loans;
  3. 7% fall in bond prices; and,
  4. 20% depreciation of the Krona

Under these harsh circumstances, Landsbanki actually remains solvent. In order to produce insolvency, the stress test would need to include a run on bank deposits, which is what ended up happening to 40 percent of Landsbanki's U.K. Icesave accounts. Keeping in mind that hindsight is 20/20, how would such a sitution have been envisioned, particularly in an economy that, at the time, was being celebrated as one of the world's most dynamic?

Applying rigorous tests to vulnerable economies is no less easy.  

While stress tests can serve as an important mechanism for banks and financial systems to protect themselves from the effects of future crises, they are too often shunned by the institutions that need them the most. These institutions fear that applying hypothetical scenarios that "imagine the unimaginable" will create a self-fulfilling prophecy, whereby the results can lead to a drop in confidence in the institution's financial health and a run on deposits. The stress test will end up creating the worst case scenario that it had imagined. 

Look ahead to much discussion about what an ideal stress test would look like, and how it can be used by those who need it most.

Comments (1) E-mail Digg Bookmark Facebook

Comments

Stress tests need to have an 'institutional' component. It is not enough to say that banks would be stable even under the most severe shock we can imagine. We should also be able to say that there is a capacity to deal with an unimaginable blow, i.e. that there are competent people and supportive rules that would enable them to respond to 'unknown unknowns'. The Swedes had the capacity to deal with the early 1990s crisis; the Eastern Europeans or East Asians less so, but still some got out of their respective late 1990s banking crises better than others.
The World Bank did a couple of pilot studies for the Global Bank Insolvency Initiative that tried looking at the preparedness of countries to respond to unpredicted banking crises. FSAPs and ROSCs also shed some light on this, although their exclusive focus on compliance with 'best practices' is misconceived. However, as the memory of the Asian crisis faded, so did these initiatives...
In short, you cannot predict the outcome of the battle by just counting guns; you may do better by knowing the generals as well.


Post a comment

Comments are moderated, and will not appear on this weblog until the author has approved them.

World Bank Financial Crisis Response | IFC Financial Crisis Response | Doing Business | Financial Systems | Remittance Prices Worldwide | PSD Blog
©2009 The World Bank Group, All Rights Reserved. Legal. Terms of Service.