A couple weeks ago the Federal Reserve announced results of the latest Comprehensive Capital Analysis and Review (CCAR), aka stress test, for large banks.  The stress scenario included 13% unemployment, a 50% drop in stock prices and a further 20% drop in housing prices.  Basically, the scenario is a severe double dip recession.  In order to “pass” the test, each bank had to estimate loan losses over a nine quarter horizon and show that they would remain in compliance with all capital requirements.  That is, capital today must be sufficient to absorb any losses due to the difficult economic scenario and still maintain a surplus relative to the capital standards.

Banks that pass the test with flying colors (i.e., with large capital surpluses) will be allowed to pay dividends or buy back stock.  Fifteen of the nineteen banks that were required to undergo the stress test passed, and four did not. Those that did not pass include Citicorp, Ally Bank, Sun Trust and Met Life and these companies will be constrained from capital payouts (like, for example, dividends) pending another examination.

As discussed in a previous blog (December 20, 2011), the Basel capital standards can be gamed.  Once a particular asset class, like Sovereign debt or AAA mortgage backed securities, is assigned a low risk weight based on assumed low risk, there are pressures to produce more and possibly higher risk versions of that asset class.  If regulators assign a risk weight that is too low for a particular loan type, banks will take advantage of the error and tend to over-weight that loan type. A better approach to assessing risk is by subjecting a strategy or portfolio to a regime of stress testing. 

There are many arguments for and against stress testing.  Opponents argue that the tests tend to be either too easy so that everyone passes, or they tend to be too difficult, so that most banks fail and no one will want to invest in banks.  Another negative argument is that banks will game the test by investing in hedge instruments that are tailored to addressing a particular stress scenario, but that may be worthless in other equally likely scenarios.

I come down on the side of favoring stress testing.  Capital requirements should focus on the ability to remain functional after a moderate to severe shock.  Stress testing is an important regulatory tool and should at minimum be used as a supplementary aid in assessing capital adequacy.  Also, stress testing can be a useful management tool even in the absence of a regulatory requirement.

The basic elements of stress testing are simple, at least in concept.  First, identify the key risk factors that drive the financial success or failure of your enterprise.  For a loan portfolio, this could include job growth, unemployment, housing prices for the bank’s footprint, as well as macro variables such as in interest rates and credit spreads.  Second, develop alternative scenarios for the key risk factors.  Third, evaluate market value or earnings sensitivity to changes in the risk factors.  One approach that has been used successfully in derivatives exchanges is to value all positions under a variety of instantaneous shocks to the risk factors.  These shocks would include declines in housing price appreciation and job growth, increases in unemployment and credit spreads, along with various shocks to the term structure of interest rates. 

The CCAR is a little different that the derivatives exchanges stress tests.  For one thing, there is only one stress scenario instead of a range of scenarios.  For another, the effect of the scenario is assessed by the impact on earnings over time, not by the instantaneous change in value.

How severe should the scenarios be?

The traditional approach to stress testing in banking is to assess the magnitude of losses that might be expected in a very extreme scenario, like the 1 in 1000 case.  The rationale for this is that large banks seek to maintain a high credit rating, like “A” or “AA,” and the one-year default rate for single or double A companies is very low.  So, the argument goes, capital levels should be sufficient to absorb losses that come along once every thousand years.  The problem with this approach is two-fold.  First, it is difficult (actually, impossible) to measure with any precision an event that is assumed to take place once every thousand years.

Second, it is unrealistic to expect senior management to seriously address this event.  After all, the likelihood of such an event coming along during the average CEO’s tenure is very small.

A better approach, I think, is to assess the performance of your balance sheet or business strategy over a range of moderately unlikely, but plausible scenarios.  If your strategy holds up well over the one in ten or one in twenty year events, then it is fairly robust.  If it does not, then senior management and the board might well agree that capital levels are too low or the strategy is too risky.      

I think the CCAR is consistent with this second objective.  As stated above, the Fed’s stress scenario in CCAR 2012 is roughly a severe double dip.  I would assess the probability as being about 1 in twenty.  It is not the 100 or 1000 year flood.

What about more extreme scenarios?  They certainly should be considered by management and the regulators, but capital levels should be based on calculations that are verifiable. Instead of trying to estimate losses in the 1 in 1000 event, and holding capital of at least that amount of loss, it is more meaningful to estimate losses in moderately unlikely scenarios and make sure that capital is adequate to absorb those losses and continue operating.  This is what the CCAR does.