The Salvation of VaR - Centered Risk Management: Mounting Expectations on Stress Testing
As a supplement to the limits of VaR, many advocated an active use of stress testing and also for it to be improved in this fi nancial crisis. This type of argument, however, has been heard in every fi nancial crises in the past. Each time, regulators and fi nancial institutions confi rmed the importance of stress testing and advocated its clearer position in the risk management frame- work. In this sense, the only difference in the current crisis from previous ones may be the exact events that are happening now, implying that adding another story to the stress scenarios might be suffi cient for many regulators and fi nancial institutions.
It is true, however, that the current fi nancial crisis highlighted more than ever before the importance of stress testing in terms of its role in supplement- ing VaR, or the need for a clearer relationship between VaR and stress test- ing in an integrated risk management framework. It is partly because many regulators and fi nancial institutions have just recently started to believe that the result of VaR (or the risk amount measured by VaR, which is often called “ economic capital ” ) was robust enough to be compared with the capital amount and consequently to be used to assess capital adequacy. The shock of this crisis puts this belief very much into question.
Once we have determined that VaR as a sole tool is not necessarily reliable enough to assess the adequacy of capital, what alternative measures could fi ll the vacuum? Stress testing has emerged in this context. This may be slightly different from the case of past crises, in which stress testing was highlighted, for example, as a tool to supplement the data shortage prob- lems of VaR for some risks in marginal markets.
Classifi cation of Stress Testing Concepts
Stress testing might be defi ned in a variety of ways depending on who is using it. So the term “ stress testing ” contains a variety of concepts. I will show some representative ones in the following.
The most common factors found in this variety of defi nitions for stress testing seems to be the method, which fi rst defi nes the stress event and then looks at the impacts of its materialization. This method can be classifi ed into smaller groups by focusing on 1) the perspective to defi ne the stress events or the type of stress events, 2) the scope of the stress events, and 3) the way of using the outcome.
For example, stress events can be captured with a macro perspective and with a micro perspective. The former is often called macro stress testing to be distinguished from the latter. This macro stress testing is often used by the authorities, including central banks, to see the impacts of specifi c macroeconomic shocks on the macroeconomy. For example, stress testing that assesses the impact of the current fi nancial crisis on the stability of the fi nancial system could be classifi ed as macro stress testing.
Meanwhile, stress testing that is conducted by ordinary fi nancial insti- tutions is to see the impact of the current fi nancial crisis on the P/L and balance sheets of the concerned fi nancial institutions. As I note later, most of the current stress testing tends to stop at seeing the fi rst direct impact.
In order to consider the repercussions of the initial impact among different entities and different markets to assess the second and third effects of the initial ones, we need to introduce an element of macro stress testing.
The scope of stress events to be captured is another yardstick to clas- sify stress testing. Stress events can generally be classifi ed into 1) events that actually occurred in the past (or slightly revised for contemporary context), 2) events based on a combination of the worst records of risk factors, 3) events represented by multiplied sigma or confi dence levels based on expert judgments, 4) hypothetical events that could be caused by macroeconomic disequilibrium, which share the same root causes of past crises, but do not look the same or even similar to past crises.
Finally, the use of stress testing can be classifi ed into two types. The fi rst is similar to contingency planning, and is designed to allow the concerned
68 POST-CRISIS RISK MANAGEMENT
institutions to react to the stress event in a smooth way once it actually occurs. Another is to assess the adequacy of capital once the stress events occur and consequently cause huge losses.
In table 4.2 , I summarize the types of stress testing based on this perspective.
What Went Wrong with Stress Testing?
One of the biggest claims about fi nancial institutions ’ stress testing that came from the regulators in this crisis concerns the stress scenario itself. In other words, they questioned why fi nancial institutions did not assume the type of stress scenarios we have already seen in the current crisis. These are events in which prices in some securitization markets dropped signifi cantly, or where the assets that had been once detached and transferred to struc- tured investment vehicles (SIVs) then returned to their own balance sheets.
Furthermore, they are also the events that explicitly consider the risks that have not been well recognized before, including market liquidity risk, sec- ond effects, or the repercussions from the initial macroeconomic shocks, and model risk concerning the risk of securitization products. According to the authorities, they have seen that the degree of stresses were generally not enough even when these risks were considered in fi nancial institutions ’ stress testing.
However, the supervisory agencies were also supposed to supervise those fi nancial institutions ’ stress testing through on - site inspections or off - site monitoring. In this sense, the claims from the authorities might not be clear even to the authorities before the crisis. In this connection, it is interesting to Table 4.2 Features of stress testing
P erspective to capture stresses Macro Micro
S cope of stresses Historical events (or slightly revised for contemporary context)
Combination of the historically worst numbers of risk factors
Multiplied sigma and confi dence levels based on expert judgments
Hypothetical events that share the root causes of disequilibrium with past crises
U se of stress testing Contingency planning
Confi rming the adequacy of capital
see the outcome of a horizontal study of stress testing, which was published by the UK FSA in October 2006. In this publication, the UK FSA clearly concluded that the degree of stresses that fi nancial institutions actually used was not enough. This indicates that some authorities had already recognized the insuffi ciency of stress scenarios in fi nancial institutions ’ stress testing but had not fi xed them in the period leading to the fi nancial crisis. This fact might lead us to conclude that the relationship between supervisors and fi nancial institutions should also be considered. I will discuss this issue further later.
In addition to the contents of stress scenarios and degree of stresses, the regulators ’ claims concentrate on the use of stress testing for manage- ment judgment. They saw that the stress testing outcomes might be used to assess risks in the risk management section, but might not be fed back to senior managers as elements to infl uence their business decisions. They also observed that the stress testing outcome did not necessarily constitute an integral or effective part of ensuring capital adequacy.
These were actually areas in which the major global fi nancial institu- tions had long boasted of their own practices as being world best, showing us how diffi cult it would be for a third party to judge whether some banks ’ risk management is better than others .